This was different.

Jean-Paul felt it even before they attended the event. The email had come in two languages – English and French – and then again as a phone call from someone the family trusted, not from an address that began with “no‑reply”.

The letter had said there would be childcare, and that the meeting would finish in time for school pickup. It had requested their attendance politely and explained that this design workshop was endeavoring to do things differently. There were travel vouchers offered, and the invitation had arrived a month before the date, not two weeks before. The letter had provided options for interpreters for anyone not comfortable in English, so Jean-Paul and Marie had asked for a French interpreter.

From the moment they walked in, Jean-Paul and Marie could see this workshop was not like the others they had previously attended.

Marie noticed the little things first. On the sign‑in table, the lanyards with their names clearly printed. Her name card said “MARIE”, and underneath, in smaller letters, her name written two ways:

Mah‑ree

/ma.ʁi/

It was a small thing, but it mattered to her. She had missed people saying her name right- the French way.

Next to the lanyards, someone had put out small flags and stickers that quietly said welcome. The ones closest to her read: “Kia ora”, “Jambo”, “Bienvenue”, “مرحبا”, “Boyeyi malamu” … she could not see the rest, but when she saw the Lingala welcome, she felt the tension start to leave her body.

A kind woman with a Red Cross lanyard bent down to four‑year‑old Chantal’s level and explained, slowly and with a smile and gestures, that there were toys and crayons in the next room, and that her parents would be just through the wall if she needed them. Giggling, Chantal left holding the woman’s hand, without a backward glance. The woman winked at Marie reassuringly as she left with her daughter.

The smell from the back of the room was familiar and new at the same time. Coffee, yes, but also mint tea, a variety of fruit, flatbread, samosas, puff-puff balls, and was that mikate?! Smells that reminded Marie of home. The food on the long table was clearly labelled with little cards: “vegan”, “vegetarian”, “gluten-free”. There was a table set aside labelled “Halal - no pork”.

“How considerate for our Muslim friends,” she thought, as her shoulders dropped a fraction more.

They had attended a government workshop before, and the title on that email had been impressive too. “Consultation on the National Refugee Strategy”, it had said. They had gone, because they had been told their voices mattered. In that room, the chairs had faced a projector screen. The plan had already been on the slides. Their comments had been requested in the last twenty minutes, after the PowerPoint presentation.

That time their job had been to “share perspectives”. They talked about childcare, learning English, qualifications that vanished on paper. People from the agency nodded and said they had captured their contributions. They had believed them, but months later, nothing in their lives felt different.

This room was set up differently from the last one. The chairs were in loose circles around the tables. There was no projector screen. On the walls, there were only large sheets of paper and clean, empty space.

Jean-Paul recognized a few faces as they found their seats. The older Sudanese man from the mosque. A young Syrian nurse he knew from the bus stop. A grandmother from Myanmar with her daughter beside her. A tall Afghan builder who always wore a hi‑vis jacket, but not today. An older Somali man he knew by sight, who had been in the country longer than any of them, he sat very straight with his hands folded, watching the officials more than the room. There were others he did not recognize. He nodded and smiled at the faces he knew.

The officials all wore branded clear name tags making it easy to tell them apart. A man in jeans with a laptop under his arm, looked like he might be “the computer person”. A woman from Palmerston North City Council he had once seen speaking at the library. Someone from Work and Income. A young lady from the Te Whatu Ora public health team and a gentleman from Community Law. Two officials from Immigration New Zealand, and a lady from Māngere Refugee Resettlement Centre who had travelled to attend. Yet they all sat around the circle, mixing with the group.

“This is the first time,” Jean-Paul thought, “that all the agencies with pieces of our story have been put in one room.”

Marie sat and ran her fingers along the nametag before putting it on.

She sighed. Last time, the invitation had felt like a summons – here she felt respected.

She turned to her husband and smiled. “This feels different, Jean-Paul. They see us. Look at how they thought about every individual person’s needs.”

He smiled back. “It looks like they finally realize we are not all the same.”

The interpreters took their places, introducing themselves one by one, then sitting directly behind the people who needed them.

When everyone was settled, a woman stood up, not behind a lectern but beside the seated group. She began with a mihi, greeting mana whenua first, then the communities in the room, then the agencies. She introduced herself by her first name – Patricia – and her whakapapa before she said her job title. Then she spoke slowly, in English, with pauses for the interpreters to follow:

“This is a government design workshop. We are here today because we realize you should have been here before anything was built – we want to correct this. So today, there is no finished plan on the wall. You are here to help us redesign service delivery for future refugees. What we design in the next months will shape how future families like yours meet this system – the practical services, and also the automated parts a computer helps decide. That is why your voice is so important.”

Jean-Paul felt Marie’s hand tighten around his under the table.

“Last time, they asked us what we thought of their plans,” she whispered. “This time, it sounds like they are asking what our lives are like, before there is a plan.”

At the next table, the man with the laptop – the one Jean-Paul had guessed was “the computer person” – nodded along as Patricia spoke. He did not open his device. After Patricia sat down, he stood up. Looking slowly around the room, he explained why he was there.

“Before, people in offices wrote the rules that decide who gets what help, and then asked you what you thought,” he said, pausing for the interpreters. “Today we are starting the other way round. We want to understand your lives first. The tools – including the computer systems that process your benefit decisions – will be built from this.”

He added that they would also be speaking with other groups – Māori, Pasifika, CALD communities, disabled people, and other vulnerable communities – so that their input would be captured in the revision of the system.

Jean-Paul glanced at Marie. “They are talking about the automated decision‑making program that the government is expanding,” he murmured. “We get to help improve the system for others.”

One of the officials from Immigration New Zealand stood up to speak. He explained that the goal for all officials attending from the different agencies was to listen carefully and capture the opinions, ideas, and issues raised in the room. Their task, he said, was to convert these into real changes that would benefit future refugees.

Jean-Paul smiled softly, finally understanding the intention in the room. It is not that Aotearoa had done nothing for them. When they arrived from the Democratic Republic of Congo, Jean-Paul and Marie felt safe, for the first time in years.

They had the skills New Zealand had a real shortage of. Jean-Paul was a psychiatrist, trained at the CNPP - the neuropsychiatric center at the University of Kinshasa - and years of experience in a busy city practice. Marie was a qualified veterinarian who had worked with livestock. Unfortunately, this did not translate into employment. Their working language had been French, with Lingala at home. English was never a language they had needed to speak. When they stepped off the plane in Auckland, every word in English was new.

At Māngere Refugee Resettlement Centre, staff and volunteers helped them through those first weeks: health checks, school enrolments, bus cards, GP registration, tenancy and bond rules, what to do in an emergency. Interpreters stood beside them in the rooms that mattered. They were housed, clothed, registered, talked through the basics of life in Aotearoa. It felt, genuinely, like being held.

When they moved on to Palmerston North, the Red Cross team and a small Congolese community picked up the thread. People in the community helped explain letters and school notices. Volunteers showed them where to buy familiar food, which buses to take, how to call the clinic. For the first months in Aotearoa, there was almost always someone they could call who understood both their languages and their new city. It was still hard, but they did not feel alone.

The safety was real. So was the gratitude. But underneath it, the obstacles were also real.

To work again as a doctor in New Zealand, Jean-Paul had learned, he would need registration with the Medical Council of New Zealand. That meant proving his qualifications and experience, passing assessments, and, crucially, showing a high level of English through an approved test before he could even enter some of the registration pathways for international medical graduates. The forms were long and intricate, full of phrases that assumed not just conversational English but exam English. He spent many evenings at the kitchen table with a French-English dictionary and his old certificates, trying to make the paperwork match the person - until he had given up in frustration.

Interestingly, psychiatry is one of the most stretched specialties in Aotearoa - about a fifth of senior posts vacant, a quarter in the Central Region where they lived, and roughly three in every five psychiatrists already working here had, like him, trained overseas. The country was short of the skill that Jean-Paul had, but it still locked him out of the practice.

A trained volunteer came once every weekend to teach both of them English. It was a genuine kindness because it helped them integrate into Kiwi life and Jean-Paul could finally chat to Chantal’s teacher. Unfortunately, no volunteer could carry him from “Kiwi English for daily life” to the exam level 7 English the Medical Council required. The help that reached his door was not the help that would open the professional one.

Marie’s path back to veterinary work was no easier. To work as a vet here, Marie’s Lubumbashi degree had to be assessed because it wasn’t on the recognized list. That meant sitting the two-part Australasian Veterinary Examination, a written paper and then a clinical exam run through Massey, before she could register with the Veterinary Council. Passing the first part would let her work only under supervision - but the very shortage that made the country want her meant the clinics had no one spare to supervise her.

Every step assumed a level of technical English she did not have yet. The registration requirements talked in a language she was still learning to navigate. The applications, the guidance notes, the exams themselves – assumed a level 7 technical English. New Zealand was advertising for livestock vets, had an experienced one right here, but she too could not practice. Even the streamlined Green List pathway, built to fly vets in from overseas, was closed to her: it required her to already be registered. So, another door closed.

For both of them, the English requirement was not just social; it was formal, written into the rules. The pathways back to medicine and veterinary practice were real, but they were also long, exam‑heavy, and built for people who started with strong English, not for people starting from zero.

At first, Marie had tried to find her way back into practice. If she could get to English classes, then to the assessments, maybe one day she could get back to her beloved profession. But Chantal was four and got only 20 free early-childhood hours a week. To put her in care any longer - for the number of hours needed to attend a course that would upgrade her English - Marie needed the childcare subsidy. And to get more than a handful of subsidized hours, the rules asked her to already be working or studying. To study, she needed the childcare. To get the childcare, she needed to already be studying. Marie was caught in a childcare-subsidy catch-22.

In the meantime, they needed to pay rent, to buy food and take care of their little girl.

So, when a supermarket job opened up, Marie did the only sensible thing the rules left her room to do: she took work that fit inside Chantal’s twenty free hours, stacking shelves while her daughter was at daycare and collecting her when the funded hours ran out. Now she was learning product codes instead of disease surveillance updates.

They took survival jobs - the only jobs they could get with their limited English. They did not resent the work. The income was welcome. The co‑workers were often kind. But every shift underlined the same quiet message: they both had the skills New Zealand desperately needed and yet somehow the jobs were out of their reach.

Jean-Paul was employed at the Warehouse doing manual labour – lifting, stacking, cleaning – work he did with care, but which had nothing to do with the years he had spent diagnosing and treating patients. Eventually he had to take on a second job, driving for Uber, to make ends meet.

Some nights, after the last fare, Jean-Paul caught himself noticing the signs he had spent a career helping others name - the flatness, the thinning sleep, the way he had started leaving calls from home unanswered. He knew exactly what to call it. He also knew that in a country this short of psychiatrists, there was nowhere to send himself. He also knew he could not afford the time off.

Sometimes Palmerston North made the loneliness easier to bear. There were other Congolese families who understood their jokes and their silences. The community spoke a language they knew. When Jean-Paul needed to talk about DRC politics, or Marie needed to reminisce about home recipes, there was usually someone who understood without explanation.

But some absences no community could fill. Marie’s mother was still in Kinshasa - a voice on a line that dropped out mid-sentence, a grandmother who missed watching her grand-daughter growing every day. And Jean-Paul carried his twin brother, the one who shared his face, still back there, still unsafe. A part of every fare he drove was promised before he earned it, sent back across the world. They could be surrounded by their own people here and still feel split in half, the other half a continent away.

A year on, the doors had still not opened. New Zealand’s only veterinary school was a few kilometers from their flat; Marie knew she did not have the papers they needed to see her skill. On the job sites, it was true: there were listings for psychiatrists and vets everywhere. However, the doors stayed shut. Some closed early, asking for New Zealand experience and full registration. Others called them in for interviews, but although their skills matched the jobs advertised, the interviews expected effortless English, and the unspoken bias was that poor English meant poor skill.

With time, something else changed. The dedicated Red Cross support - the guidance of the caseworker who not only spoke French and Lingala but also helped them navigate their new home - had to stop. The support had always been time limited. It ran its course over the first eighteen months, thinning as it went, and then it ended. Jean-Paul and Marie were in the mainstream system now, relying on MSD for income support and job-seeking help, like everyone else. The trouble was that “like everyone else” assumed a footing they did not have.

MSD saw them as work‑ready, sometimes even “over‑qualified”, for the kinds of low‑wage jobs advertised on its website – and simultaneously invisible for the roles that matched their training. Automated decision‑making systems helped process their payments and reapplications, but those systems did not ask – and could not know – that Jean-Paul had once sat with patients in crisis, or that Marie understood biosecurity risks better than most people in Aotearoa.

New Zealand offered theoretical opportunities but not access. The vacancies were there. The pathways were there. The rules, exams, and English thresholds that sat between them and those opportunities were there too – but there was a disconnect between the promise and the lived reality.

While the support lasted, when a confusing letter arrived, there had been someone to sit at the table and read it line by line. The letters still came - the same official English, the same deadline buried in the third paragraph - but now they were met with dread and a French-to-English dictionary. Their English had come a long way in their time here, but they still struggled.

That was the shape of the vacuum they carried into this new workshop: safe, but under‑used; grateful, but tired; highly skilled in exactly the areas New Zealand said it needed, yet working in jobs that did not touch those skills; shifted from a world of named supporters into a world of automated decisions and generic categories.

Patricia the workshop guide stood up again.

“Today,” she said, “we are not asking you to comment on a finished strategy. We are asking you to help shape the rules before they are written into forms, letters, or computer systems. The letters that arrive in your mailbox or as decisions that affects your life – we want to understand whether they need to be redesigned to improve the outcomes. Your life stories are the starting point.”

The interpreters moved a beat behind her. Heads around the room nodded at slightly different moments, like a soft echo.

She glanced around the circle, checking that everyone had understood and that no one had been left behind.

Marie felt something unfamiliar: No one was rushing them through a list of bullet points. No one had said, “We only have ten minutes for this part.” Here they were all listening for their answers.

Jean-Paul listened to the words “…before they are written into forms” and remembered that letter that had arrived in complicated English, with a deadline hidden in the third paragraph. He wondered, for the first time, if he might suggest that all important government letters be made available in languages other than English, some did but he wished more could. Maybe, he thought, they could even use safe translation tools to make that possible. He knew that Canada, Singapore and the United Kingdom were using this technology.

Patricia walked over to one of the blank sheets taped to the wall.

“In a moment,” she said, “we are going to start with your stories. Not ‘feedback’, not ‘comments’ – your stories of what happens when you meet this system: at Work and Income, at the hospital, at school, with Immigration. We will listen first. Then the agencies here will ask: if we considered your experiences, what would we need to change, and what would we keep?”

She put the marker down on the table instead of writing anything yet. The page stayed empty, waiting.

The process was explained slowly. First, Patricia said, each person would have time on their own to write or dictate their answers to a small set of questions, with their interpreters and staff on hand to help. Then they would talk in small groups, to see what was the same and what was different in their stories. Only after that would anything go up on the walls.

“Everyone here,” she said, “has the right to be heard in their own words. So, we will capture your words first, before we try to turn them into themes.”

Staff moved quietly around the room, handing each person a sheet with four questions printed on it:

• What is working for you — in health, housing, work, education, community?

• What is not working? (What things are affecting the quality of your life)

• What did we miss?

• What should change before this goes wider?

Each question was written in clear English, and the interpreters read them out slowly in other languages. Where people preferred to speak rather than write, interpreters wrote their answers down exactly as they said them, in their own words, checking back to make sure nothing had been lost.

Patricia walked slowly around the circle, acknowledging everyone, grateful for the variety of stories in the room. A Syrian man who had spent months trying to get his overseas social work qualifications recognized. A young nurse from Burundi working nights in aged care while she chased qualifications that would be recognized here. A Burmese young man who did most of the talking for his family at the clinic. A former refugee youth worker who knew the school system from both sides. Different ages, different professions, different countries – each writing their stories on the clipboards they had been given.

In the corner opposite Jean-Paul sat a dignified older Somali man, he did not pick up the marker. He looked at Patricia and spoke through his interpreter. “I have sat in rooms like this before,” he said. “Good rooms. Kind people. We talked. They wrote our words down.” He let the words rest. “Then nothing changed.” He folded his hands again. “I will tell you about my life. But forgive me if I do not clap yet. I have clapped before.” Patricia nodded at his words with respect. No one rushed to reassure him. His interpreter simply wrote down what he had said, word for word, and thanked him.

Marie watched the care being taken. No one was paraphrasing their words into “key challenges” or “themes” yet. When she said, “The letters from the government make me feel stupid, even when I know I am not,” the facilitator wrote exactly that on the page.

Jean-Paul’s first instinct was to answer: “Warehouse laborer / Uber driver (two jobs).” For months, that was the version of himself the system seemed able to recognize. Then he saw the next line: “What work did you do before you came to New Zealand?” For a moment his eyes misted.

“Doctor,” he said softly to his interpreter, then asked him to add, “Psychiatrist.”

Beside him, Marie wrote “veterinarian” and after a pause, explained to her interpreter: “I worked on animal disease and food safety – stopping sickness before it reached people.”

The interpreter paused at Marie’s words. For the first time, he seemed to grasp the full significance of Marie’s expertise: not simply caring for animals, but helping stop sickness before it crossed into homes, markets, and communities. In a country with persistent veterinary shortages and strong reliance on biosecurity, that kind of skill was not marginal; it was urgently relevant.

At the far side of the room, the man in jeans with the laptop – the engineer – sat with his own clipboard in front of him. He was not typing. He was writing, in the same thick black marker as everyone else. When one of the participants looked over and asked what he was doing, he smiled and said, “I am learning how you experience the system - what it looks like now from your side - and what you would prefer it to be instead, before I touch a keyboard.”

When the other near-by participants looked up at his comment in surprise, he shrugged. “Usually, they hand me the rules and I build them. This time I get to listen to the people I actually build these systems for.”

The room settled into a busy buzz. The staff were not just collecting stories; they were listening for definitions. For how harm sounded in the mouths of the people who felt it. For what dignity looked like in a Work and Income office. For which parts of the hospital journey made people feel safe, and which parts made them feel small.

As Jean-Paul went to re-fill his mint tea at the tea urn, he caught the edge of a conversation that wasn’t meant for him. Two of the officials talking to each other, paper cups in hand. “It’s the right way to do it,” one was saying, quietly. The other frowned “But think about it, we have to run this again in Auckland, Christchurch, then Wellington, the time and money this takes is just….”

Jean-Paul didn’t wait to hear the rest. He didn’t want the doubt in the sentence to settle in his body. Thankfully, his basic English would not have allowed deeper comprehension anyway.

When the individual pages were complete, they turned to small-group conversation. At each table, Patricia asked the same four questions again, out loud this time:

• What is working?

• What is not?

• What did we miss?

• What should change before this goes wider?

People pointed to their own sheets and to each other’s. The Burmese young man spoke about how visual health leaflets and pictograms had helped his family. The Syrian young girl talked about her hotel manager who changed the roster so she could get to her English classes on time. Jean-Paul talked about the English proficiency required to take the New Zealand Registration Examination (NZREX) needed as an internationally trained doctor. He explained that this sat between his experience and any chance of practicing again. Marie talked about reading the veterinary registration guidance and feeling as if she were standing at the bottom of a mountain she could not climb.

The facilitators did not rush them. They repeated back what they had heard – not in policy language, but in the same words people had used – and checked if they had it right. When someone suggested a change, it went down on paper with their name beside it, not folded into a vague “community feedback” box. The respect and recognition for their lived experiences were obvious.

Before they closed the workshop, Patricia made one more promise. In six months, she said, they would be called back. Each agency returning with reports on what they had tried to change, everyone in the room would be invited to say what had worked, what had not, and what still needed to happen. This was not a one-off consultation. This was the beginning of a relationship.

Jean-Paul looked at the older Somali man across the room. His dignified face did not give much away, but he had watched Patricia’s face as she spoke. When she was done, the old man held her gaze a moment longer than everyone else – and gave her one slow nod. A man marking a date he intended to keep. In six months, that look said, I will be here. And I will see for myself.

That evening, Jean-Paul and Marie did not leave with a strategy document under their arm. But with the unfamiliar, fragile feeling that someone had finally written their experience into a blueprint.

Now, I have sat in both kinds of room: the one where the plan is already on the slides, and the one that opens with a blank wall and a question. The difference is not comfort, or kindness, or good intentions. It is the difference between “Do you agree?” and “How is it for you?“ The difference between asking people to approve a blueprint and trusting them to draw one. Between being a subject of the system and an author of it. Aotearoa is at the cross-roads – it is time to choose which room we build for the vulnerable among us.

Jean-Paul and Marie are invented. The struggles they live with are not. On 1 July, New Zealand widens automated decision-making across more of its welfare system. Most of the families those rules will affect were never in the room where the rules were written. The only question now is whether we correct the room before expanding the system even further - or after the harm is done.

_{Denise Barlow-Byarugaba is a Human Systems specialist and the founder of The Room Institute. She practices through the lens of Ubuntu-informed AI governance - a framework that redesigns systems that make decisions about people, from policy to automated decisions, with the communities that carry the cost of our failures. She draws on almost 30 years in community development, walking alongside Māori, Pasifika, and CALD communities across Aotearoa and East Africa. She has worked on both sides of the glass: the frontline, where she has seen what poorly designed systems do to people, and inside the build, where she has seen how they are made. She partners with government agencies and NGOs to redesign those systems so that the vulnerable are protected, as welfare policies and emerging technologies land in real lives.}

Thanks for reading The Room! Subscribe for free to receive new posts and support my work.

_{Image was created with Canva AI}

Most people in Wellington walk past her. Not because they are cruel, but because they have to keep moving. It is cold, the southerly cuts down the street, and there are always emails to answer, buses to catch, children to pick up.

Rachel looks like many of the young women who gather on Cuba Street. Beanie, hoodie, backpack. At a glance, she could be waiting for a friend. If you look twice, you see the difference. The way she watches every passer-by without really looking at anyone. The way she holds her bag as if it is the only proof she owns anything at all.

Rachel is only twenty-one. She is alone. And this winter, she is sleeping on the streets.

There is kindness in her day. A soup kitchen, a social supermarket and free store gave her the dignity of choices, a drop-in that provided free health care and basic needs and another that provided a place to clean your laundry and get a warm shower. On paper, there is a lot of help. Someone kind will check that she has a warm blanket when the night is particularly bitter. In that sense, the city has not abandoned her.

But the help that would actually get her off the street – emergency housing, stable income, a door she can close and lock – sits behind processes she struggles to navigate.

Every process Rachel needs to walk through assumes a version of her that does not exist.

The forms assume she has a fixed address to write down. The appointments assume she can keep track of dates and times when she has nowhere safe to keep a phone charged and no quiet place to think. The interviews assume she can sit across from a stranger and tell her story calmly without her throat closing up from shame.

They assume she knows and understands these processes exist in the first place.

The agencies ask the same question: “Is this person eligible?” Eligibility is a question about her. It puts her on trial. Does she qualify? Does she have the right documents? Did she attend the appointment? Did she cause her current situation? How long has she been without employment? Is this a reflection on her character? Every one of these questions is a test she must pass. Each one feels like a judgement. She is expected to answer them in her traumatised hypervigilant state. How could she when she can’t think beyond where she is going to sleep tonight?

Trauma takes the part of you that plans and trusts and tells your story to a stranger at a counter. That part goes first. Addiction, for those who carry it, is rarely the start of the trouble; it is the thing that quiets the noise in your head. So, the cruelty is built right into the shape of the system: the very conditions that most qualify a person for help - the fear, the disorganisation, the inability to hold a thread - are the same conditions that disqualify her from operating the help. The more she needs it, the less she qualifies for the help. The door stays open, and those behind the door say she walked past it, and choose not to engage. They record this as her choice.

The truth is, Rachel was not an addict when she arrived on the street. The addiction came later, as an adaptation, not a cause. It came with the wet clothes, the constant stress and the sleepless nights. It came with the constant calculations of where to sit so she is visible enough to be safe but invisible enough not to be moved on. It came with the need to blunt the cold and soften the edges of the fear.

People see the substance use and assume it is the reason she is here. She knows it is one of the only ways she is managing to stay here at all.

Yes, we do offer help. But let us look closely at what we offer.

The truth is we offer fragments of help, kept in separate buildings, held by separate agencies that do not speak to one another. Housing is one ministry. Income is another. Mental health is a third. Addiction, a fourth. Refugee resettlement, a fifth — each with its own building, its own criteria, its own waitlist, its own version of the eligible person it is waiting for.

So Rachel is not facing one door. She is standing at the mouth of a corridor of them, and each one opens onto a different office that knows only its own slice of her. To get what she needs she must walk to many offices, sit in front of different case workers, fill in different forms, tell her story again at every desk, become a slightly different applicant each time, and carry the information from one application to the next herself. We are asking her to be the connecting web between agencies that will not connect - to hold the whole picture together inside her own exhausted mind. A system operating in silos that has never bothered to hold together on its own.

Yes, it is easy to say that Wellington has services. Work and Income. Housing support. Housing First and outreach teams. Soup kitchens, day centres, shelters, drop-ins. The doors do exist.

From where Rachel is sitting, the question sounds different: do these services reach her? – in this state, with this brain and this body in winter – or do they wait for her to reach them?

Right now, the system still asks the same question over and over: “Is this person eligible?” Rachel is living proof that it is the wrong question. She will still be sitting on the bench on Cuba street next week.

Asking the question “Is this person eligible?” sounds responsible. It sounds like careful stewardship of scarce resources. But it is the wrong question.

The better question - the only one that would actually find the people who truly need our help - is a question about us: Can she reach our service?

Let’s consider: There is emergency housing, in theory, for people with nowhere safe to sleep. There is Work and Income emergency housing process with its forms, assessments, and waitlists. There is Housing First that requires she prove that she has been on the streets long-term before she qualifies, Wellington City Mission that requires documentation of alternative avenues, Kāinga Ora’s long and involved application process that relies on MSD for its referrals. I do notice the good mahi and the reprieve offered by these agencies, but I want us to acknowledge the hurdles that an already stressed mind would experience.

Tragically most of these services are not only oversubscribed, they also only offer transitional housing - a welcome relief, but not a permanent solution. Atareira, Hutt Valley Community Housing, Tuatahi Centre, and Wellington Homeless Women’s Trust, are some of the agencies that offer this temporary reprieve.

Credit needs to be given to the outreach workers whose job is to find people on the street and walk alongside them for as long as it takes. There are community agencies like Downtown Community Ministry, the Soup Kitchen, the night shelters, women’s refuges, and churches that open their doors when everyone else has gone home.

Taken together, they look like a well-coordinated network, providing a safety net to catch those who slip through – and most times they do. However, the danger is that we could walk away feeling - “we are doing all we can”.

The numbers tell a harder story.

The latest national homelessness insights report, produced by Te Tūāpapa Kura Kāinga – Ministry of Housing and Urban Development, shows tens of thousands of people across Aotearoa living in severe housing deprivation – including people in cars, garages, boarding houses, and emergency accommodation. The official social housing register still sits at 19,704 applicants waiting for a public house as of 31^st March 2026.

In Wellington, frontline services like the Soup Kitchen and Orange Sky report more people turning up for kai, showers, and support, not fewer, and warn that homelessness is rising faster than the city is growing. Downtown Community Ministry’s own count of rough sleepers climbed from 114 to 141 between 2024 and 2025, a 25% increase. So, evidence suggests that the rough sleeping number keeps climbing every day. After falling since 2023, the social housing register is climbing again - up 2.1% in the past year.

At the same time, access to emergency help is getting tighter. After changes to emergency housing rules in 2024, people can be declined if officials judge that they “caused or contributed” to their own need. One analysis found that decline rates for emergency housing applications skyrocketed from just over 3% in 2023 to nearly 34% - more than one in three - by 2025. Advocates and food banks report a similar pattern in hardship support: more people asking for help, more food grant applications turned down. This means that the vulnerable already fear a real possibility of a decline even before they apply.

So yes, the services exist. The system can show you the doors. It points to a register that has fallen from over 25,000 since 2023, as progress - but its own officials warn that the number drops not because the need was met, but because people give up applying. On paper, some indicators are even improving, the government declares its 75% reduction target met five years ahead of schedule as success. It can point to increases in social housing supply and call that progress.

However, for Rachel on the bench on Cuba street, the experience is different. She sees a list of phone numbers and addresses she cannot hold in her head. She sees offices that want more answers than she can give. She sees documentation that will be requested and forms that she has to complete. She sees waiting rooms where she is not sure which version of her story is safe to tell. It’s the survival instinct; she always feels it in her body – so what she tells a government agency is often different from what she can tell an outreach worker.

Rachel hears the government is building more homes for people like her, but the truth is – the hurdles around them are too high for her to climb over. Her heart carries one grateful truth – “Thank God for outreach workers!” - they are the heroes in her story. She sees them doing their best – but with more people to reach than hours in the day, she knows they just touch the tip of the iceberg.

But they are the ones who still respect her mana, who refuse to let her disappear. Instead of waiting behind a desk for her to arrive, they go to her. DCM’s outreach team, Toro Atu, walks the streets to meet people who are rough sleeping where they are, treating that first contact not as an assessment but as the beginning of a relationship. Through the Aro Mai Housing First collaboration, they work to move people into a permanent home first, then wrap support around them so the tenancy holds - and along the way they sit with her to get her ID, untangle her entitlements, and complete the forms she cannot face alone. They navigate the corridor on her behalf, rather than asking her exhausted mind to navigate it for them.

Kaupapa Māori providers have held this for decades. He Herenga Kura has walked alongside the vulnerable for over fifty years, grounded in Manaakitanga (dignity and well-being), Kotahitanga (collective responsibility), Rangatiratanga (autonomy), and Whanaungatanga (strengthening relationships), with housing and support designed specifically for people like her, offered to Māori and non-Māori alike. Their model does not begin by asking whether she qualifies. It begins by asking what she needs to thrive, and it holds the whole of her - her wellbeing, her connections, her future - rather than a single slice. This is what mana-enhancing practice means: help that restores dignity instead of demanding she earn it.

Other agencies are the exact opposite. They split her needs and never see her as whole. Each agency can point to the piece it is responsible for; her accommodation, her health, her budget, her employment. None of them is responsible for whether the whole picture adds up to safety for her. No one holds the whole of her. At a distance, it is easy to believe the system is neutral. A form here, a checklist there, a series of criteria applied the same way to everyone. This is what it looks like when one’s life is handled by silos. On paper, it looks fair, it sounds reasonable and responsible. Rachel seems to be surrounded by services that solve her problems.

But the system was not built for her. It was built for a phantom applicant: someone who has a phone and data, a fixed address, stable mental health, a dry place to keep her documents, a clear head that remembers the days of the week, and the confidence to sit across from a case manager and explain their life in neat, linear sentences. Someone who can come back next week, and the week after, if the process asks for it. Someone whose pain does not leak out sideways as fear and frustration take over. Someone for whom trauma, abuse or violence on the street, never turns into a missed appointment. That person almost never exists in real life - and it’s certainly not a twenty-one-year-old on Cuba Street, in winter, on her sixth month living outside on Wellington streets.

To these agencies Rachel is one of the invisible young people on the streets, unless she walks through their door. This is a quiet danger. This is how we miss her. Community Housing Aotearoa has warned that many people experiencing homelessness do not engage with MSD, HUD or similar agencies at all - so accurate numbers are hard to pin down because administrative records will undercount and misrepresent the very people it most needs to see. Leaning harder on agency data than on the census, CHA cautions, risks producing statistics that obscure the true scale of homelessness. For Māori. For women. For youth. For her.

Yes, the system keeps asking the same question: “Is this person eligible?” And every time she fails to meet the model it was designed for, it records that as her failing, not its own. It sees her as hard to reach. From where Rachel is sitting, it is the system that is hard to reach. This is especially relevant now, in this new era of automated decisions that the government is expanding into. Today, the system makes a judgement on who is considered “Priority One”, who “didn’t keep the appointment” and therefore no longer qualifies, who doesn’t have the right documentation and so “gets sanctioned“ These rules are increasingly being written into software. The phantom applicant becomes code. The moralising gate becomes a decision tree. The same tests she cannot pass while unwell are turned into conditions an algorithm can check at speed.

What policy makers call “decision support” and “advancement” can easily become decision replacement. Today, it will misread just one young woman on Cuba Street. It gets it wrong because the data points it uses are wrong. Tomorrow, it will get it wrong for thousands, because automated decisions scale. Consider the harm of this.

Even now as Rachel sits on this bench, her world narrows even further. Parliament is considering the Summary Offences (Move-On Orders) Amendment Bill, which would give police the power to move her on. There is a call to make her even more invisible.

_{Image was created with Canva AI}

Interestingly, so many agencies spoke up against the harm the move-on orders might cause. Police advised that this was not the best way to address poverty and homelessness, plainly saying they do not support criminalising rough sleeping and non-aggressive begging. Oranga Tamariki opposed applying the orders to young people, noting that existing models of care were more appropriate. Officials at Housing and Urban Development warned against the bill, the Ministry of Justice preferred the status quo. Corrections warned that it would likely need major investment in further prison infrastructure and the Treasury questioned the Government’s claim - that only six people a year would be jailed - calling it unlikely to withstand scrutiny. Several other agencies voiced their concerns about how the basic rights of young people and women sleeping rough would be violated by the move-on orders. Putting them at greater risk of harm, abuse, and exploitation if moved on from well-lit central spaces, especially at night.

The sharpest warning is about children. A Public Health Expert Briefing, published June 2026 by researchers from the universities of Otago and Waikato and Pūrangakura, sets out that the orders would allow children as young as fourteen to be detained, fined, or criminally charged for behaviours that are simply the face of homelessness - sleeping in a doorway, sitting with a sign asking for money. Almost half of all people experiencing homelessness in Aotearoa are under twenty-five – this statistic has Rachel’s face on it. Unfortunately, the state has not built the alternative: Oranga Tamariki has only 154 supported accommodation places for roughly 1,618 young people aging out of care each year, two-thirds of them Māori. There is nowhere to move on to. This is not a gap the researchers invented - in its Kāinga Kore report, the Waitangi Tribunal found the Crown had already breached the principles of active protection, equity, and good government, naming the lack of support for homeless rangatahi as an ongoing breach. This bill compounds it.

Amnesty International Aotearoa New Zealand has called for homes not handcuffs, warning that the powers are discriminatory and - given the ongoing effects of colonisation - will fall disproportionately on Māori. As their Executive Director Jacqui Dillon put it, “when a person is forced to move, particularly one who may have no other place to go, you remove their right to exist within the community.” They note the bitter irony that while England and Wales are repealing their Vagrancy Act, Aotearoa appears to be leaning back toward the Victorian era.

The Attorney-General and Housing Minister, Chris Bishop, filed a Section 7 report warning that key provisions of the bill are inconsistent with the New Zealand Bill of Rights Act - specifically the begging and rough-sleeping grounds. Community Housing Aotearoa, together with the Māori housing advocate Te Matapihi, formally opposed the move-on powers and called instead for housing-led action.

The bill has drawn unusually broad opposition, from opposition parties, almost every agency that is involved with the homeless, homelessness advocates, Māori organisations and various city councils.

The opposition to the bill is clear. The advice was given. The advice was not heeded.

Actually, the advice was not simply unheeded. It was overridden: the Justice Minister batted away his own officials’ warnings, saying the government stands by its approach.

Now, let’s consider what a move-on order actually does to her.

It does not produce a bed, an income, or a door she can lock. It produces a reason to disappear. A woman whose whole problem is that no one sees her is now, by law, deliberately moved on so she cannot be seen - further from the only safety she knows – the soup kitchen that knows her, the outreach worker who was building trust, deeper into the alleys where danger waits and caring eyes cannot follow. Move-on does not solve homelessness. It relocates it. And in relocating her, it strips an already invisible woman of the only protection she once had.

An Ubuntu-informed governance lens does not allow her to become invisible. It moves her back into the frame, where she belongs.

1. It designs with the person on the street; they get to define what help looks like.

2. It does not begin by asking whether the person can fit into the system, it focuses on simplifying paperwork, rigid criteria, and the processes that would make access to emergency housing overwhelming.

3. It asks whether it reaches the people who are unable to come to the service – understanding that people without shelter are often highly mobile, making it difficult to maintain contact with standard case management systems.

4. It assumes trauma or distrust before it assumes non-compliance – and so focusses on building relationship and gaining trust.

5. It asks, “what have we missed” and writes in revision at every stage – with a goal to continually close gaps it discovers.

In practice, that would mean fewer corridors of separate doors. Complete care loops that circle the whole person. And listening first before any decision is made. It would mean asking, at every stage: Who is not here? Why are they missing? Is the data we have capturing everyone? Whose story is being read as fault instead of harm?

It would mean measuring success, not by how many services exist, or how many are completing the application, but by how many people like her are no longer sleeping rough.

We still have people slipping through the gaps not because we don’t have support systems, but because the way we chose to shape the resources we offer is - application-based, siloed, conditional, counting the people who walk through the door and no one else. Rachel is not a gap between services to be tidied up. She is the most honest feedback the system has. If our doors do not reach her, it is not she who is failing the test. It is the test that needs to change.

But fortunately, her story isn’t over yet. This is not settled. The Summary Offences (Move-on Orders) Amendment bill is before the Justice Select Committee now, and submissions are open until 2 July 2026. If this article has moved you to action, that is the room you can still walk into - you can make a submission. This is your opportunity to speak for the people who are slowly becoming invisible. It is possible for YOU to give them a voice.

A note on ‘Rachel’: the young woman in this piece is not one person. She is a composite - every detail of her circumstances is true to lives I have sat with over twenty years of frontline practice, but she herself is an anonymised figure.

I have written her this way deliberately. The people I have spent my life serving are too often turned into a type, a risk score, a row in a dataset - flattened into someone the system can process. Drawing her as a composite lets me protect the real people behind her while keeping faith with the one thing this whole piece is about: that behind every number is a particular human being the system was supposed to see.

_{Denise Barlow-Byarugaba is a Human Systems specialist and the founder of The Room Institute. She practices through the lens of Ubuntu-informed AI governance - a framework that redesigns systems that make decisions about people, from policy to automated decisions, with the communities that carry the cost of our failures. She draws on almost 30 years in community development, walking alongside Māori, Pasifika, and CALD communities across Aotearoa and East Africa. She has worked on both sides of the glass: the frontline, where she has seen what poorly designed systems do to people, and inside the build, where she has seen how they are made. She partners with government agencies and NGOs to redesign those systems so that the vulnerable are protected, as welfare policies and emerging technologies land in real lives.}

In an African village, a newly widowed woman faced a devastating crisis. Her late husband’s brothers, driven by greed, moved to seize her land, crops, and home, claiming ancestral right. Left with young children and stripped of her livelihood, the widow did not suffer in isolation. She brought her grievance directly to the village square. The community elders immediately halted the village routine. The talking drum was sounded to gather the entire collective in the village square. In this circle, the greedy relatives did not meet immediate banishment or a sterile, detached sentence. Instead, the elders invoked the foundational pillar of Ubuntu - the truth that echoes throughout Africa in different tongues: Umntu ngumntu ngabantu (Zulu), literally “a person is a person through other persons.” The call of I am because you are.

The brothers were forced to face the widow, their community, and the memory of their ancestors in the open air. For hours, the elders and villagers engaged in intense, transparent dialogue. They did not just demand the return of the land; they spoke to the brothers, helping them recognize that by starving the widow, they were severing their own humanity and tearing a hole in the fabric of the entire village. True justice was achieved only when the brothers felt the weight of their collective shame, took accountability, and pledged to protect her.

The resolution was sealed not with punishment, but with a communal feast where the family ate from the same pot, physically mending the broken kinship. In the worldview of Ubuntu, justice was never about punitive isolation; it was the meticulous restoration of relational harmony. The Ubuntu values are evident: relational accountability, collective decision-making, transparency, and restoration are the core pillars of this community.

Centuries later, in Aotearoa New Zealand, another widow sits at her kitchen table, facing a terrifyingly similar crisis. Now a solo mother trying to survive on a single income, she applies for urgent social welfare support to keep a roof over her children’s heads. However, instead of a supportive community circle, she is met by a cold, automated algorithmic system. A predictive AI risk-scoring model flags an administrative discrepancy in her historic data, misinterprets her frequent change of address (her continuous search for affordable accommodation) as potential fraud, and automatically terminates her benefits.

Here there is no village square for her to turn to. She cannot reason with the machine, nor can she easily find the human bureaucrats hidden behind the proprietary “black box” code. In this modern landscape, the prevailing model of AI governance leaves her harmed. And she has to formally appeal to have this decision overturned. All of this takes energy, presence of mind and grit – all of which she has in short supply. The system will view her crisis as an isolated technical glitch. It offers her only a bureaucratic appeals process, treating the tech provider with nothing more than mild regulatory oversight.

The algorithm, much like the greedy brothers of the past, extracts from the vulnerable while remaining entirely unaccountable to the community it harms.

I write from Ubuntu, the relational worldview I was raised in, and I bring it into conversation with whanaungatanga and Māori Data Sovereignty - not as a lens laid over them, but as a kindred worldview meeting them as equals. Ubuntu and te ao Māori reach the same relational truth from different shores. Standing together, on the constitutional ground of Te Tiriti o Waitangi, they both call for a radical paradigm shift in how AI governance works in Aotearoa New Zealand.

Through this lens, algorithmic harm is not a minor data error but a collective injury to communal well-being. Repairing it asks three shifts of our governance in:

I. how transparent these systems are,

II. who holds real power over their design,

III. and who owns the outcomes they produce.

Naming three shifts is the easy part. Anyone can call for transparency, shared power, and community ownership. The harder question - the one that matters in practice - is what design actually delivers them. That is where my Ubuntu-informed framework comes in: five elements that run the length of a system’s life, from who sits in the room when it is first imagined to who keeps revising it once it is live. The three shifts are the what. The five elements are the how.

To understand how Ubuntu applies to modern machine learning, I will demonstrate that it is not merely a vague moral sentiment, but a rigorous, sophisticated framework of indigenous jurisprudence. Unlike Western legal traditions rooted in Enlightenment individualism - where rights are weaponised to protect personal or corporate autonomy - Ubuntu operates on a relational ontology. “I am because you are” is the paramount collective fabric that unites all.

Consequently, within an Ubuntu legal framework, a breach of justice is not categorised as an infraction against an abstract state or an isolated data contract. Instead, it is recognised as a profound rupture in communal harmony.

This philosophy holds that when a vulnerability occurs - such as the systemic neglect of a solo mother by an algorithmic system - the responsibility to rectify the harm falls upon the entire collective, including the creators of the system. A relational digital framework moves beyond the Western model of retributive, closed-door legal battles and financial settlements. It calls instead for a “palaver” – an exhaustive, public, consensus-driven dialogue where the hidden mechanisms of harm are exposed to the community, and the offender is brought to look upon the human cost of their actions. The ultimate objective is always the meticulous restoration of relational balance and accountability, ensuring that the marginalised are reintegrated and the collective structure is made whole once again. No perfection is expected, but an attitude and process of continuous revision and correction within the system.

In Aotearoa New Zealand, Ubuntu meets a worldview of equal depth that reached the same relational truth on its own whenua - whanaungatanga, expressed today through Māori Data Sovereignty. Whanaungatanga holds - just as Ubuntu does - that identity and well-being are inseparable from kinship networks (whānau, hapū, and iwi). So, when an automated social welfare system cuts off a vulnerable mother, the harm is never individual; it ripples outward, destabilising the wider community network.

This relational reality is precisely why the Māori Data Sovereignty Network, Te Mana Raraunga, asserts that data is not an inanimate, exploitable resource, but a taonga (treasure) imbued with whakapapa (genealogy and connection).

Where Ubuntu and whanaungatanga meet, AI governance finds a relational architecture strong enough to be grounded explicitly in the Te Tiriti o Waitangi principle of Rangatiratanga (self-determination and sovereignty).

When data is recognised as possessing whakapapa, it cannot be stripped of its cultural context, processed by opaque algorithms, or weaponised against the vulnerable. Just as the village elders used ancestral lineage to re-establish the widow’s right to her land, the Crown’s governance of technology must shift to honour the collective. It must treat digital ecosystems as living extensions of the community’s mana, establishing a legal framework where technology is held accountable to the collective well-being of the people it was built to serve.

Translating the transparent, open-air justice of the Ubuntu village square and the marae ātea into New Zealand digital policy requires an end to the corporate “black box” model. Currently, proprietary AI systems used in the public sector hide behind intellectual property laws, preventing citizens from understanding how life-altering decisions – such as the automation of social welfare – are made. To honour the values of Manaakitanga (upholding mutual respect and dignity) and Ubuntu, New Zealand AI governance should require Explainable AI (XAI) and open-source architectures across public service algorithms.

In the traditional village square, the widow’s grievance was laid bare in the open air, ensuring that no hidden agendas or secret decrees could dictate her fate. If a machine makes a determination that affects human survival today, the state should provide an accessible, digital “palaver” — a transparent forum where the code, data inputs, and logic weights are laid bare to the affected community and their advocates. Affected individuals should have the right to look their digital regulators in the eye and contest the algorithmic logic in a language accessible to non-technologists. Democratising access to automated decision-making ensures that technology enhances human mana rather than diminishing it through hidden, unaccountable biases.

Realising this relational governance model requires a fundamental shift in who owns outcomes. This maps directly onto the fiercely contested issue of Māori Data Sovereignty in Aotearoa today. Recall that in the village square the brothers themselves had to admit their mistake and personally own the execution of the solution; the resolution was never a decree handed down by the elders. Accountability lived with those responsible – the brothers. The justice recourse was observed and acknowledged by the whole community.

This is the philosophical equivalent of Data Sovereignty — the right of a people to own, control, and safeguard their own information, assets, and stories, rather than surrendering that power to an external corporate or state entity. In modern AI governance, the dominant frameworks remove this accountability; when an algorithm causes harm, tech giants pay a sterile financial fine to the state, maintaining absolute control over their code and leaving the harmed community disempowered.

To honour both Ubuntu and the principles of Te Mana Raraunga, New Zealand policy should embed true Data Sovereignty. Tech developers and government agencies cannot simply outsource their failures to regulatory bodies. They should be required to return ownership of the data, the algorithms, and the digital solutions themselves to the iwi, hapū, and communities affected.

Technology creators and agencies should remain actively accountable to the communities whose digital assets and stories they use. When indigenous collectives hold their own data sovereignty, they move from passive subjects of automated bias to the rightful authors of their own technological futures.

This is not abstract. Māori are already raising digital marae. Te Kāhui Raraunga’s Te Pā Tūwatawata - a decentralised, Māori-owned data network built on open-source technology and grounded in tikanga - was designed so that sovereignty does not stop at the server door: the data lives in Māori hands all the way down to the infrastructure itself. The digital marae is not a metaphor I am proposing. It is already being built.

In the same spirit, Ubuntu-informed AI governance starts with the people who live the consequences, not with those who merely have an interest in the issue. It begins by asking those closest to the harm to define it in their own words. It does not flatten iwi, hapū, whānau, Pasifika communities, disabled communities, CALD communities, children, or refugees into one generic category.

That means designing with a clean slate, not a finished blueprint. It means involving the people bearing the harm in defining the red flags, the risks, and the remedies from the beginning, and several times during the necessary continuing revisions.

That is the difference between a system that studies vulnerable people and a system designed with them. One asks them to explain the damage after it has happened. The other trusts them to help define the design through their lived experience.

Aotearoa’s Algorithm Charter - the cross-government commitment signed in 2020 on how agencies should use algorithms - was drafted with good intentions: a pledge to serve with transparency and accountability. Part of that commitment is responding to opportunities to correct the harm that continues. Calling in the right people to define that harm is the first step. This is true whanaungatanga - where AI systems are not treated as isolated tools, but as active participants in a web of interconnected relationship between humans, data, and technology, particularly in the landscape of Māori Data Sovereignty.

As community advocates, our role is to hold state entities accountable to whanaungatanga within AI governance. To do this, we build a robust architecture of five design elements that turn those three shifts into practice:

1. Move from consultation to genuine co-governance

Whanaungatanga cannot be achieved via a checklist or a one-off snapshot of the issue. In practice, it requires moving past surface-level consultation toward true partnership. This means involving Māori digital experts, iwi (tribes), and hapū (sub-tribes), along with other vulnerable groups (Pasifika, CALD, refugees, disabled people) at the foundational architecture stage of an AI system - long before an algorithm is deployed. Collaborative frameworks, such as the Māori AI Governance Framework developed by Te Kāhui Raraunga, show how public and private sectors can operationalise these shared decision-making processes.

My Ubuntu-informed governance framework calls this - The Room – similar to the village square where the talking drum called the whole community together, and everyone had representation. This is where the design and architecture of the system is built and continuously corrected: the place where the stories and needs of all the community members are heard and correctly captured. Who gets into the room anchors the architecture of the system.

The talking drum — still sounded today to call the community to gather.

2. Re-centering accountability and obligations

Under whanaungatanga, data is understood as a living extension of whakapapa (genealogy and connection). Governed AI systems must shift from managing data points to balancing collective rights and accountabilities. There is an active obligation to ensure that the data fed into AI models does not harm, misrepresent, or exploit the communities from which it was gathered.

My Ubuntu-informed governance framework calls this - Data Points – making sure the system is built to ask the right questions and measure the right things, with harm defined by those who’ve lived it. Today, it is the vulnerable - the ones who have experienced the harm and can therefore define it accurately - who should be consulted during the design. This ensures that the issues are named correctly, the right questions are asked, and the correct data points are obtained. In short, they ensure the village - the system - never goes astray. This is how we make sure the information fed to the system, and the questions asked, allow safety, correctly see the vulnerable, respect the individual and the community they represent.

3. Long-term relational impact over short-term transactions

Western AI deployment often focuses on immediate efficiency, speed, and cost reduction. All of these are necessary and have their place. However, whanaungatanga requires assessing the long-term, intergenerational impact of an automated system on a community’s well-being. It asks governance boards to consider:

● How will this AI affect our relationship with the people we serve over the next generation?

● Does this system alienate or disconnect people, or does it foster healthier societal bonds?

● Are we designing with the people’s values, needs, and protections in place?

● What frameworks, charters, and laws are the guardrails protecting those who will use the system?

This is true accountability and responsibility to the vulnerable people we serve.

My Ubuntu-informed governance framework calls this - The Frameworks - the wisdom that holds the ethics and moral code as safety guardrails for all. Today, laws, frameworks, and charters keep the automated system’s architecture and design safe. In the Ubuntu context, it was the wise men and women of the village who held the moral north star that directed the whole community. Today, the frameworks we use within the system become these wise village elders.

4. Seeing the whole person, not the presenting problem

Under whanaungatanga, a person is never a single, separable case. They are whole only in context - their relationships, their community, their history, the web of people and circumstances that surround them. You cannot understand someone - let alone support them - by looking only at the problem they put in front of you. Yet this is exactly what automated systems do: they take the presenting issue and run in a straight line to a decision, capturing a few narrow data fields along the way and calling it the full picture.

Consider the single mother suspected of fraud because of her changing address. The system flags the changed addresses but not that she keeps moving to cheaper suburbs each time. It does not ask why she moved. It does not see that her husband died ten months ago, does not realise she now lives across town from her old community, and never connects any of this to her growing depression. It never asks about her inconsistent paycheck, whether she has a support system, whether other agencies are involved, or who depends on her. Those unasked questions were not extras - they were the rest of her story. The layers the system never reached were the very layers that would have revealed her true need. It saw a woman who had changed addresses seven times in ten months. It didn’t see the struggling widow with three young children, trying desperately to hold on.

This is the difference between an open loop and a closed one. An open loop begins and ends with the symptom. A closed loop goes all the way around the person - their relationships, their resources, their history, their dependents, the dynamic of the community they live within - and gathers the whole, nuanced picture before any decision is made. People are not static and they are not simple; a system built around human lives has to be able to see them in full. The whole person is never flattened or lumped into one homogenous category but seen as the nuanced individual they are before they are funnelled toward the specific help they need.

My Ubuntu-informed governance framework calls this - The Loop. The intention to capture the whole person, never the fragment. In the village square, when someone brought their trouble before the elders, the elders never weighed the presenting complaint alone. They considered the whole person and everything that surrounded them - their whānau, their standing, their history, the people who depended on them, the relationships that held them up. The full picture was gathered before a word of judgment was spoken, because the village understood that no one’s story can be told from a single flat situation. Everyone comes with layers. To close the loop is to refuse to flatten a person into the one question they came to ask.

5. Mitigating algorithmic bias through community context

Algorithms frequently inherit the historical and systemic biases present in societal data. Whanaungatanga addresses this by establishing checks to ensure AI interprets data within its proper cultural and community context. Under this principle, the “knowledge providers” – the people whose data trains the system - and the tech creators maintain an ongoing relationship to audit, refine, and check the AI through a built-in review cycle. The goal is to reduce data drift and prevent algorithmic discrimination, bias and misrepresentation.

My Ubuntu-informed governance framework calls this - Continuous Amplification. Within Ubuntu, the village square was always open to all; it was understood that problems never stopped. Life is always dynamic and the landscape is always changing - so periodic village meetings, where all were heard, were built into community life to accommodate this. This is old wisdom, and it is especially relevant to our lives today, and to the AI landscape. Our needs, our politics, our world are constantly changing. Continuous revision is a necessity, so the communities carrying the most harm should be periodically invited back into the room for the revisions that will inevitably be required. AI has the potential to do enormous good if calibrated right – but it can cause serious, compounding harm if audits and corrections are ignored.

Admittedly, automated decision-making within the social sector is new territory for Aotearoa, and we need to acknowledge this as we all navigate the new landscape. Some government agencies are doing more than others, and that is recognised and acknowledged – especially since these governance guidelines are voluntary rather than enforced. But we would not be doing our job as community advocates if we stayed silent in the face of continuing and compounding harm.

Whanaungatanga is a sound base for agencies to check how well they are doing — an effective internal scale for measuring whether they are providing the minimum of care within AI governance.

As I watch the growing community of advocates for the vulnerable gather around this issue, I smile. It reminds me of a proverb my grandmother loved to say:

“Kamu kamu gwe muganda” – which literally translates as: “One by one makes a bundle.”

The deeper meaning is although one reed is weak, a bundle of reeds is strong enough to hold up a large structure. It is a reminder that big challenges are taken on step by step, and that through collaboration and community effort, even the most impossible tasks become manageable. All your voices are needed, because together we can create change.

Ultimately, the path to truly responsible AI governance in Aotearoa New Zealand asks for an infusion of indigenous wisdom into a largely Western, individualist legal tradition. When Ubuntu and whanaungatanga meet – two relational worldviews reaching the same truth from different shores – are operationalized through Māori Data Sovereignty principles, their meeting offers a robust architecture for a healthier digital future. The lessons of the village square show that algorithmic harms are not isolated code glitches or mere financial liabilities. They are the digital equivalent of the greedy brothers’ actions — direct tears in the communal fabric that threaten collective well-being.

This relational model is realised by adopting all three of the shifts named at the start:

I. First, the state can restore the open-air transparency of the village square and the marae ātea by opening corporate “black boxes” and replacing them with accessible, transparent digital palavers.

II. Second, the Crown can move beyond consultation to genuine consensus-driven co-design, giving affected communities real authority over the systems that shape their lives.

III. Finally, the system can move beyond state-centric ownership of outcomes toward true Data Sovereignty, asking tech developers to carry genuine accountability and to return control of digital solutions to the communities affected.

In an era where our social welfare is increasingly shaped by automated decision-making, anchoring policy in indigenous wisdom protects us as a collective. It is the preservation of our humanity and the protection of communal mana – this foundation should always guide the systems we create to serve the vulnerable among us.

_{Denise Barlow-Byarugaba is a Human Systems specialist and the founder of The Room Institute. She practices through the lens of Ubuntu-informed AI governance - a framework that redesigns systems that make decisions about people, from policy to automated decisions, with the communities that carry the cost of our failures. She draws on almost 30 years in community development, walking alongside Māori, Pasifika, and CALD communities across Aotearoa and East Africa. She has worked on both sides of the glass: the frontline, where she has seen what poorly designed systems do to people, and inside the build, where she has seen how they are made. She partners with government agencies and NGOs to redesign those systems so that the vulnerable are protected, as welfare policies and emerging technologies land in real lives.}

Image: Canva Stock

Maria was an immigrant, so she had moved into his house. Eleven years, a marriage and a child later, everything had changed. The relationship had turned sour.

On paper, she was still married. The house was in his name. He had a good income. As far as the database was concerned, she belonged to a household that owned its home and was not in urgent need.

In reality, he had thrown her out of that house and moved his new partner in. She was raising their nine-year-old alone and trying to find somewhere safe and stable to live. There was no legal divorce yet, no formal separation order - just a very real gap between the life the system saw and the life she was actually living. It was messy.

Maria was a New Zealand citizen now, so she applied for public housing support. She assumed the system would see what any human could see: a mother and child with nowhere secure to go. She did everything the system asked of her. Instead, she received a letter saying she did not qualify. No priority on the waitlist. No explanation of the reasoning. Just “no”.

In this version of the story, she knew enough to ask for the reasons. She knew enough about the process, enough to challenge the system, to write back and say: “Tell me why.” Many people would not. Many people never do.

So let me tell the story as it usually goes.

Freda is in exactly the same situation as Maria. She receives the same refusal letter, and although she had lived in New Zealand for thirteen years, she still didn’t know how to navigate the welfare system. She does not know that she can appeal. She does not know what data the system is using or how to challenge the assumptions behind it. She is already exhausted from the separation, the humble bank account, the hunt for cheap accommodation, the exhaustion she has to hide from her son, and the shame of asking for help. The letter lands like a full stop.

So, she sends her ten-year-old to school and reports to work – all the while living out of her car. Every day she tries to hide the stress and turmoil in her life. Desperate to keep up appearances. But she knows it is not sustainable. The winter cold bites different in a car.

Freda spends her lunch breaks on the phone, on hold, trying to get through to someone who will help. Eventually, in desperation, she walks into an organisation she once donated to, never thinking she would one day be on the receiving end - the Salvation Army - and there she is connected to a local advocate. Only then does she learn that there is an appeals process, that the decision can be challenged, that the “no” on the letter doesn’t have to be the end of the story.

By that point, the harm is already done. Her son’s asthma is acting up – his health compromised by living in the car. Freda no longer trusts the system she once believed in. The system that did not see her in her hour of need now asks her to step back into it – but does she have the stamina to appeal?

The question to ask the system in this scenario is: Who defines harm?

The forms Maria and Freda filled out asked the wrong questions. They did not capture their whole stories. And there was no human on the other end to capture the nuances. So, the answer to their request for public housing was ‘no’, and no red flag was raised.

That is what it looks like when protection only activates at the point of trouble with the design, not at the point of design itself. The system is reactive not remedial, so harm is inevitable.

Thankfully, the Algorithm Charter 2020 does not only shape how agencies think about Māori partnership under its Treaty commitment; it also sets out the only broad cross-government commitment that covers everyone else affected by algorithmic systems.

However, let us examine this promise. The Charter promises that agencies will retain human oversight and provide avenues for challenging decisions influenced by algorithms. On paper, that sounds reassuring. In practice, it raises harder questions.

· What happens when a decision is challenged?

· How is it reviewed, and by whom?

· How is the decision overturned?

· On what grounds can it be overturned?

· How is bias identified and corrected, not only in a single case but in the system itself?

· How are tikanga, cultural competence, and conscience built into the oversight process in ways that communities can trust?

· Are there guardrails?

These questions are the specification the Charter should contain and doesn’t. The truth is this - You cannot retain what was never held. The Charter promises to keep humans in control of a process they were never in control of. The algorithm was always doing the heavy lifting.

If the system is built on the wrong data points at the start, then “human oversight” becomes reactive rather than protective. It activates after the harm, not before it. It assumes that the person harmed understands the process, trusts the system enough to challenge it, and has the time, energy, literacy, and support to act. Those are not small assumptions. For many vulnerable people, they are enormous ones.

Freda eventually does get emergency housing - but only after push back, after she has mentally stilled herself for a fight. After her trust in the system is broken and her child is sick. This is not a small cost, especially because she is just one of many.

A protection that only works if the harmed person knows the rules, trusts the referee, and can afford to keep playing is not much of a protection at all. This protection lets down the people it is meant to protect.

The Charter lets us down on the people commitment too.

Commitment 3: “People”, says agencies should focus on people by “identifying and actively engaging with people, communities and groups who have an interest in algorithms, and consulting with those impacted by their use.”

At first glance, that sounds reasonable. On closer reading, it is thinner than it looks.

This is the only place in the Charter where Pasifika communities, CALD communities, disabled people, children, refugees, and other groups affected by algorithmic decisions really appear as a collective concern.

This commitment has three clear weaknesses:

The first issue is the ceiling of the engagement. The Charter sets consultation as the highest guaranteed relationship with impacted people. Sherry Arnstein’s classic ‘Ladder of Citizen Participation’ places consultation in the middle band of tokenism, while partnership sits higher up in the band of citizen power. Consultation can be a legitimate step, but if it is not linked to shared decision-making power, it remains what Arnstein called a “window-dressing ritual”. A rubber stamp that only ticks the box. So, when the Charter promises consultation to those most affected, it is not promising co-design. It is not promising shared authority. It is promising a tokenism-rung relationship to the people it most needs to protect.

The second one sits in the verbs- consultation vs engagement. Agencies are asked to ‘actively engage’ with those who “have an interest” in algorithms, and to “consult” those who are “impacted” by their use. The people the system is actually meant to support - the ones who bear the harm - get the weaker verb, those with an “interest” - the experts, researchers, industry voices, observers looking in from the outside - are invited into the room to ‘actively engage’. The hierarchy is upside down. Those who live with the consequences are given consultation - the window-dressing token. Those who study, manage, or debate the issue are given active engagement.

The third one is timing. Commitment 3 is silent on when engagement and consultation happen. Consult after the blueprint is drawn, and the agency has still technically met the wording. Consult once the architecture is fixed, and the obligation is still, on paper, fulfilled. The Charter does not require a clean slate. It does not require communities to be there when the harm is first defined or when the system logic is first set.

And again, underneath these three weaknesses sits a more fundamental question: who decides who counts as “impacted” in the first place?

If a system claims to focus on people by consulting those who are impacted, then someone has to define what the harm is, where it sits, and who carries it. But if the people who live with that harm are not in the room when those definitions are formed, then the harm is still being described from the outside in.

It is like designing a house you have never lived in. You can sketch the floor plan. You can choose where the walls go. You can decide how many bedrooms there should be. But if you have never had to move through that house as a single mother with a child, or as a disabled tenant, or as a refugee family navigating systems in a second language, then you do not know what rooms are actually needed.

That is the weakness in Commitment 3. It has the feel of a problem being studied and documented, not solved. Those “with an interest” are actively engaged. Those who live the consequences are consulted.

This is why I keep returning to the difference between open loop and closed loop design.

An open loop observes the problem from the outside. It studies the issue, measures the outputs, waits for complaints, and responds when something goes visibly wrong. It treats people as data points and cases to be managed.

A closed loop does something different. It keeps going back to the people who live the consequences and asks: what did we miss? What has changed? What are we not seeing from here? It understands that people, situations, and harms are dynamic. They evolve. Any system built around human lives has to be able to evolve with them. The people ARE the data. Without them, you don’t understand the picture.

That is one of the deepest weaknesses in the charter. It does not require continuous revision with the people affected. It does not require agencies to return to communities once the tool is deployed and ask whether the assumptions still hold. It does not require a loop to be closed.

There is also a quiet demotion risk within the Algorithm Charter 2020. Commitment 2, “Partnership”, is Treaty-specific and sits in a different category from Commitment 3, “People”. That distinction is key. If agencies start treating Māori as simply another community to consult under Commitment 3, rather than as Treaty partners under Commitment 2, then Māori standing is quietly demoted from partnership to stakeholder status. The Treaty relationship begins to dissolve into the general people bucket. What should be a distinct commitment to shared authority becomes ordinary consultation.

This is not a minor wording issue. It changes the power structure of the whole document.

Freda sleeping in her car should not have to become an expert in appeals before the system sees her. Her child should not have to experience the harm before human oversight wakes up. And the people most affected by algorithmic decisions should not be consulted as an afterthought while outsiders are actively engaged as the main conversation partners.

This is why my Ubuntu-informed AI governance lens keeps returning to the same principle: Communities are not the subject of the system. They are the source of it.

_{Denise Barlow-Byarugaba is a Human Systems specialist and the founder of The Room Institute. She practices through the lens of Ubuntu-informed AI governance - a framework that redesigns systems that make decisions about people, from policy to automated decisions, with the communities that carry the cost of our failures. She draws on almost 30 years in community development, walking alongside Māori, Pasifika, and CALD communities across Aotearoa and East Africa. She has worked on both sides of the glass: the frontline, where she has seen what poorly designed systems do to people, and inside the build, where she has seen how they are made. She partners with government agencies and NGOs to redesign those systems so that the vulnerable are protected, as welfare policies and emerging technologies land in real lives.}

My first introduction to community development was in Uganda, my home country. It showed me what true community participation looks like, and it has shaped more than 20 years of practice since then.

We went into a village where there was very little: no school, no proper road system, no health facility, and limited economic opportunity. Like many development teams, we gathered the community and asked what they needed. But we also did something important: we spoke separately with the women, the men, and the young people. That was deliberate, because not everyone speaks freely in a mixed room.

What they said at first seemed to pull in different directions. The women said they needed water. The men said they needed income. The young people said they needed something to do. In many systems, that would have been the end of the conversation - someone would have chosen the “most urgent” need, written the report, and called it consultation.

We did not do that.

Instead, we took those answers back to the community and asked them to work through the tension themselves. We did not arrive with a blueprint. We arrived with a clean slate. What they came up with surprised us: the community knew we couldn’t provide everything, so they asked for two things: a well and a lorry.

The well made immediate sense. The lorry did not look like the obvious answer. It was not the neat, textbook solution that outsiders would usually prioritize. But it came from the community, so it was trusted. And in time, that choice changed everything. The lorry carried produce to bigger towns, opened up income, reduced dependence on middlemen, and helped the community build its own future. The well freed up women’s time, which created space for craft, trade, and enterprise. Young people gained skills by handling the trade, the marketing, and the well repairs. The community became stronger from the inside out.

That experience became the foundation of the Ubuntu‑informed lens I work within: a way of designing systems that begins with community knowledge, honors lived experience, and refuses the habit of building solutions for people without first building them with people.

That lesson matters now more than ever, because the same mistake keeps showing up in AI governance: communities are invited to respond to systems they did not help design.

In Aotearoa New Zealand, the Algorithm Charter 2020 is not the only document shaping AI governance, but it is a foundational ethical guide for the public service. It stands as a cross‑government commitment, signaling how agencies should develop and use algorithms - including AI - fairly, ethically, and transparently. It asks agencies to manage risks, mitigate bias, maintain human oversight, and embed Te Ao Māori perspectives into their processes. The question, then, is not whether the words sound good, but whether the Charter actually requires the shared authority those words imply.

Too often, vulnerable communities are not co‑designing the systems that shape their lives. They are being consulted after the fact, asked to comment on a blueprint already drawn somewhere else. And when that happens, consultation becomes a performance of inclusion rather than a transfer of power.

In practice, this is especially visible in how the room treats Māori. Māori are often the only community with a seat at the table - and they hold it by Treaty right, not by grace. Yet even that seat is frequently narrowed to a “perspective” to be embedded, rather than authority to be shared. Everyone else- Pasifika, disabled people, CALD communities, children, and others who bear the system’s effects - is not offered the seat at all. They are flattened into a vague category of “users” or “stakeholders”, their realities are treated as secondary even when the harm is shared. So the document performs two reductions at once: a partner narrowed to a perspective, and a person narrowed to a category.

Imagine this. A government agency invites an iwi trust to comment on a new decision‑making tool that will affect whānau in their community. The agency says it wants Māori input, and the meeting is framed as a partnership. But by the time the trust is brought in, the system’s purpose, risk categories, and core logic have already been set. The iwi representatives are asked for feedback on wording, communication, and “cultural considerations” - not on the design itself. The invitation sounds collaborative, but the architecture of the system is already fixed. The iwi representatives’ input is only within narrow boundaries. Their role is to improve a system they did not shape, and their presence is then used to signal legitimacy. That is not co‑design. That is rubber‑stamping with a Māori face on it.

That is the crack.

If AI-automated systems in social services are deployed in Aotearoa New Zealand without being designed by vulnerable communities - even if they are rubber‑stamped by Māori consultation - then those systems were designed for the vulnerable, not by the vulnerable. This distinction matters. It is the difference between partnership and permission, between co‑design and consultation, between shared authority and symbolic presence. The people bearing the harm are not involved in defining it; the agency defines the risk, the red flag, and the solution. The gatekeeper and the gatekept are the same entity - the agency builds the system, then rates its own risk for it.

Fortunately, the Algorithm Charter 2020 gives us the right language to hold it accountable. Its Partnership commitment asks agencies to “deliver clear public benefit through Treaty commitments” by “embedding a Te Ao Māori perspective in the development and use of algorithms consistent with the principles of the Treaty of Waitangi.” That is the standard it sets for itself - so let us test it against the Treaty principles.

• Partnership requires shared authority at the design table. If Māori are only brought in after the system is already built, then there is no partnership - only permission. A perspective is not the same as presence, and presence is not the same as power. You cannot partner on a blueprint you were never there to draw.

• Participation requires voice with weight. Consultation alone does not meet that standard if the community can speak only within narrow boundaries, after the core logic has been fixed. And when all other vulnerable communities are flattened into a generic category, participation becomes even thinner. Attendance is not participation. Rubber‑stamping is not voice.

• Protection requires the protected to help define the harm. If the gatekeeper decides the risk, the harm, and the safeguards without those most affected in the room, then protection becomes supervision. It may look careful, but it is not relational, and it is not accountable to lived experience.

• Tino rangatiratanga / Increasing authority - increasingly recognized as a fourth strand alongside the settled three - requires authority to grow. It cannot grow from absence, nor can it be built on consultation that leaves decision-making power unchanged. If communities are not in the room from the start, authority remains where it has always been - with the institution. Nothing increases from zero if zero is all that is offered. .

That is the problem with the Charter on its own terms. It speaks the language of Treaty alignment, but it does not require the conditions that make Treaty principles real. It allows good intentions. It permits excellent practice in some places. But it compels nothing. Remove the goodwill, and the floor is bare.

Some agencies do go further. Oranga Tamariki, for example, has Māori advisory roles and pockets of genuine co‑design. The truth is that this happens despite the Charter, not because of it. The good practice depends on individual agency goodwill, not on anything the document requires. The Charter permits excellence; it compels nothing.

My practice - Ubuntu‑informed AI governance - would do things differently.

Ubuntu says, “I am because we are.” It refuses the idea that people arrive at their circumstances in isolation. It insists that context, relationships, history, and community are not background noise; they are the data. An Ubuntu‑informed approach to AI governance takes the lesson of the village square seriously: you do not arrive with the answer, you arrive with a clean slate.

Applied to AI, that looks very different from the Charter’s model:

• You begin with separate listening. Different groups - iwi, hapū, whānau, Pasifika communities, disabled communities, CALD communities, refugees, children and young people — are never flattened into one homogenous definition; each needs a safe space to define the problem in their own words. Each suffers the harm uniquely.

• You surface apparent contradictions and do not rush to resolve them. The tension between “we need income” and “we need protection” might be the place the real design work has to happen. Listen to solve, not just to tick a participation box.

• You allow solutions to emerge from the people most affected. The “lorry” in an AI context might be a locally governed data infrastructure, a community‑controlled decision rule, or a veto power - something no external ‘expert’ would have designed alone.

• You ensure community ownership of outcomes — continually. The system, the data, and the safeguards are not simply explained to the community; they are shaped, refined, and governed by them. A system that allows for continuous revision - by the people, for the people - acknowledges that people, situations, and dynamics change. To create a robust architecture, you need to plan for this in the design.

In other words, the people bearing the harm are involved in defining it - and in designing the tools meant to address it.

That is the difference between the Charter and the Ubuntu‑informed AI governance model. One treats communities as a source of legitimacy. The other treats them as the source of the system.

_{Denise Barlow-Byarugaba is a Human Systems specialist and the founder of The Room Institute. She practices through the lens of Ubuntu-informed AI governance - a framework that redesigns systems that make decisions about people, from policy to automated decisions, with the communities that carry the cost of our failures. She draws on almost 30 years in community development, walking alongside Māori, Pasifika, and CALD communities across Aotearoa and East Africa. She has worked on both sides of the glass: the frontline, where she has seen what poorly designed systems do to people, and inside the build, where she has seen how they are made. She partners with government agencies and NGOs to redesign those systems so that the vulnerable are protected, as welfare policies and emerging technologies land in real lives.}

Why I’m opening this room

For more than twenty years, my work has been about rooms.

Not boardrooms or conference rooms - though I’ve been in plenty of those. I mean the rooms where decisions get made about people’s lives. The case meeting. The eligibility interview. The multi-agency meetings where client cases are discussed. The panel that decides whether a family gets help or gets turned away. I’ve spent two decades in those rooms, working alongside Māori, Pasifika and CALD whānau here in Aotearoa, and earlier in East Africa, and I learned one thing above all else:

The most important person in the room is often the one who isn’t there.

The whānau being discussed but not present. The person whose file is open on the table while they sit outside, or aren’t invited at all. Their absence doesn’t make them matter less. It makes the decision worse - because a room that can’t see someone can’t truly serve them.

I’m telling you this because the rooms are changing.

More and more, the decisions that shape vulnerable people’s lives aren’t being made by a person in a room at all. They’re being made by automated systems - algorithms deciding who gets flagged, who gets help, who gets seen, who gets reduced to a risk score. The room has become a piece of software. Lines of code. And the same question I’ve been asking for twenty years matters more now than it ever has: who’s in the room when this gets built? And who got left out? Whose voice was not captured?

That question is why I’ve spent the last while moving into AI ethics and governance. People sometimes treat that as a reinvention - a frontline community worker becoming something else. It isn’t. It’s the same work. The same conviction - that the people most affected by a system are the least likely to have shaped it, and the most likely to be harmed when it goes wrong. I haven’t changed my cause. I’ve followed it into the place where the decisions are now being made.

I will be honest, it’s only in the last year- 2025, that I’ve stepped deliberately into AI ethics - close enough to the start of the road to still remember what it looked like from the outside. And the more I learned, the more I realised ethics was only one arm of it. It was the arm that was asking is this right? But I realized that this was just one half of the whole. It wasn’t enough on its own; you also have to ask who is accountable, who decides, who can be made to answer when it goes wrong? That second question is governance - the structures and rules around a system, not just its intentions - and it’s the other arm I’m still actively navigating. The two need each other. Ethics without governance is good intentions with nothing to hold them in place. Governance without ethics is process with no conscience. We need both.

I’ve also done something a little unusual for someone coming from where I come from: I’ve gone inside the machinery. Part of my work right now is with the AI labs themselves - helping train and shape the AI systems - and it has taught me how they actually function under the hood, not how they’re described in the brochure. Sitting on both sides - the frontline, where I’ve watched what these automated systems do to people, and the inside, where I can see how they’re built — is what lets me grapple with automated decisions honestly. It’s the whole picture. And it’s what I now carry into every room: not just the practitioner’s question of who gets harmed, but a working sense of where, in the machine, the harm gets made.

This is the room I’m opening here. A room that holds two realities at once: a candid look at how these systems fail the people they’re meant to serve, and the language of AI governance to explore what we do about it — and to invite more voices into the mahi. The two sides of one coin.

The Room is where I’ll think out loud about all of this - the ethics, the harm, the design, the strange new shape of old problems. Some of it will be about the systems, and the robust frameworks that audit them, or honest explorations of how we better serve those who may fall through the cracks. Most of it, really, will be about people: the ones I’ve met over these twenty years whose lives were never as simple as the form that tried to capture them. A woman keeping a folder of screenshots to prove her own case to a system that should have held that record for her. A man whose missing paperwork was treated as his failure, when it was really the system asking the wrong question. The immigrant who is struggling to navigate the automated voice at the end of the phoneline, that only serves to trigger old traumas. You’ll meet some of them in the writing to come. My aim is for you to realize that the statistics, are real people, with real lives. I put faces to the numbers.

My practice is grounded in Ubuntu - I am because we are. It’s an African philosophy, and it has been the quiet foundation under everything I’ve done, in every room I’ve walked into, whether or not I named it out loud. It means a person is never an isolated data point. They are held in a web of relationships, history, and context, and you cannot understand them - or serve them, or build a system around them - if you simplify their complexity and flatten it away.

So that’s what this room is for. Thinking, in the open, about how we keep the human whole as these systems decide more and more of our lives. About who belongs in the room. About how to capture the nuances that could be easily missed. About making sure the people I’ve spent my life serving don’t slip through cracks that no one seems to see.

If any of that resonates with you, pull up a chair. There’s room.

Denise

_{Denise Barlow-Byarugaba is a Human Systems specialist and the founder of The Room Institute. She practices through the lens of Ubuntu-informed AI governance - a framework that redesigns systems that make decisions about people, from policy to automated decisions, with the communities that carry the cost of our failures. She draws on almost 30 years in community development, walking alongside Māori, Pasifika, and CALD communities across Aotearoa and East Africa. She has worked on both sides of the glass: the frontline, where she has seen what poorly designed systems do to people, and inside the build, where she has seen how they are made. She partners with government agencies and NGOs to redesign those systems so that the vulnerable are protected, as welfare policies and emerging technologies land in real lives.}

Why subscribe?

Subscribe to get full access to any new articles I publish and publication archives.

Stay up-to-date

Never miss a piece - every new post is sent directly to your email inbox.

Join the conversation

Be part of a community of people who are a part of this important mahi. We need more voices. Participate in the comments section, share your opinions or ask about the work.

I welcome all contributions.