The Broker Between the Scientists and the State

SIGLA (System for Intelligent Growth and Learner Anthropometry) is a tool that screens kindergarten pupils for malnutrition, using computer vision to read a child’s height and weight from a single photo—work that teachers otherwise do by hand, child by child. The stakes are high. One in five children under five in the Philippines is stunted, and the consequences reach directly into the classroom: chronic malnutrition in early childhood impairs brain development during the window when it matters most, leaving lasting deficits in memory, attention and cognitive capacity. Stunted children get sick more often, miss more school, and are more likely to drop out—compounding the neurological damage with reduced access to education itself.

Karl Satinitigan is the education specialist the Philippines team worked with on EdTech Hub’s Ministry of Education AI Challenge. The Department of Education already had the technical talent: its in-house AI team, E-CAIR (Education Centre for AI Research), which had built SIGLA. The project needed a translator—not of language but of politics. Coming to AI from public policy, Karl was able to help instill confidence with the ministries and read a government that’s eager to make AI more visible. 

The conversation below is drawn from two interviews, lightly edited and condensed for clarity.

Over six months and across six countries, six ministries of education joined EdTech Hub’s Ministry of Education AI Challenge. Each Ministry worked with embedded AI experts to build their first AI prototypes. Meet the AI Experts introduces the people behind those tools: who they are, how they think, and what it takes to build AI a ministry of education can actually use. For this second profile, we’re featuring Karl Satinitigan in the Philippines.

Let’s start with you. What prepared you for a role like this—working at the intersection of AI and government?

A few things came together. My background is in public policy where I’ve spent seven years working for a national policymaker, and through most of last year I was a consultant to the Congressional Commission on Education, which was this major reform push for a school system that is in real crisis.

The pandemic compounded an already severe malnutrition and stunting crisis in the Philippines, and with the public acutely aware of the stakes, there’s a short window to act. And then you have AI arriving at the same time. In a lot of those conversations AI was seen as a threat. So to have the chance to show it could be part of the solution was exciting.

How would you describe what you actually do, to someone who’s never heard of this role?

Honestly, this is my daily challenge when I meet new people. On paper I’m an innovation manager. For innovation to happen you have to manage stakeholders, work out who holds the decisions that let people actually innovate. But the role changes every day. There’s no fixed list of tasks.

My role is actually closest to a product manager in a startup. I have to understand what the data scientists are doing well enough to communicate it to the officials and policymakers we need buy-in from. And I have to understand what that other side needs, what motivates them, what frustrates them, so I can tell the scientists that while their work is great, here’s what we need to prioritise given limited time.

In short, I’m a broker. There’s a lot of information flying around on both sides, and part of my job is working out what each side actually needs to hear, because if the officials get overwhelmed they switch off, and if the scientists feel unheard they get frustrated.

One of my quieter missions was simply to keep that team working on this. Experts like them are incredibly hard to find, and very easy to lose to a private company that would pay them more. I wanted them to see that the opportunity cost of working in government is real, but so is what you can change from inside it.

Early on, the pilot got stuck, and it took time to work out why. What was going on? And how did you get it moving again?

One piece of context: ECAIR is relatively new to the education ministry, having begun life in a different department. So at the beginning, a lot of people in the ministry simply didn’t know them yet. It wasn’t about anyone being difficult, they just weren’t sure whether this team was one of them or a group from outside.

The clearest sign was the pilot itself. The team needed to roll out SIGLA across ten schools to begin the initial data collection. By around November the team had only managed one out of ten.

They had champions, but at the very top, the cabinet-secretary level. Lower down, as a new team they ran into processes and approvals they didn’t yet understand, because they come from the private sector and academia, where things work differently. 

So in the first months our main job was to mediate between ECAIR and the rest of the ministry, and to work out who was who: who actually controls the energy in a room, who can nudge a particular office. Through my networks, we found other champions who could help us understand where the delays were really coming from.

A few things helped to get the pilot back in gear. Workshops and stakeholder consultations to build the relationships, and then EdTech Hub being able to shoulder some of the resources so the data collection could move ahead while the usual approvals ran their course. That let us pull off a second round of data collection before the end of the year, and by March we’d finished all ten schools.

If you were briefing someone stepping into a brokering role like yours tomorrow, what would you tell them?

Take time at the start just to listen, before you dive into everything that needs doing. In my first weeks, the most useful thing I did was sit in on meetings between ECAIR and the rest of the ministry and watch how people talked to each other, and who actually controlled the room. A lot of that is information you can’t get any other way, because no one will ever tell you that this supervisor is loved by her colleagues, or that this staff member quietly decides what gets prioritised, not because of her title but because she’s been there longest.

Brokering is as much about what you don’t say as what you do. Especially in the Philippines, face-saving and power dynamics matter enormously. Understanding that is what tells you where to push and where to pull.

Then the mood flipped. Suddenly there was pressure to go national. How did you read that, and how did you keep the team grounded?

The ministry enjoyed the results so much that they wanted to take it to the entire public school system, around twenty thousand schools, straight away. The pressure came right from the top. In January, ECAIR got to present to the president directly, and he was excited.

There’s a learning moment in that for anyone in AI policy: alongside the education crisis, the national government was also looking for a success story, with a lot of corruption scandals around at the time. There are moments in public service where, ready or not, you’re expected to contribute something that shows progress, and you can use that momentum to push your agenda. But you have to temper it with the reality of doing this at scale.

To their credit, ECAIR had always wanted an intermediate phase. We helped by setting up a session with one of the challenge’s AI experts to talk through the realities of scaling: the cost of securely storing visual information, techniques like federated learning, the fact that the language they were building in could get expensive to maintain at scale. Some of it ECAIR already knew, some was new, but it gave them the confidence to say: we’re not ready to go nationwide. The fact that we’d already struggled just to do ten schools made the argument stronger. A good model isn’t the whole job. So they pushed for a hundred schools first.

So what does the path forward look like now?

Right now ECAIR is fine-tuning the model with everything the ten schools taught us, and drafting the budget for the hundred. Going from ten to a hundred brings new costs, especially orienting teachers and parents, and the transport and logistics of reaching schools well outside Metro Manila. We’ve offered to help make that budget plan workable.

The aim is to run the hundred-school data collection within the first term of the new school calendar, perhaps by August, though the timeline isn’t fixed yet. The team will go through the bureau’s own data to choose the hundred schools, prioritising the most severe malnutrition and the widest spread of geography, especially remote areas. The ministry already does manual screening at the start of every year, so this slots into a moment that already exists.

Did you see the people you worked with inside the ministry change how they think about AI?

Yes, and in a specific way. Most teachers, principals and parents in the pilot were already using AI—ChatGPT or Canva—for their own work. What shifted was a bigger idea: that AI isn’t just a tool for individual tasks, but something that can reshape systems. It could, for instance, reduce the time teachers spend hand-recording students’ height and weight data—and, at scale, use that same data to direct food resources more effectively to where malnutrition and stunting are worst.

They liked the obvious wins, getting results faster, spending less time measuring. But the more interesting conversations were about using data as evidence. I remember a principal in December, and then a health officer in charge of the feeding programme around March, who said this could change how she deploys her staff. That’s the moment I felt people’s sense of what AI could do for them genuinely expanded, beyond what they normally associate it with.

Last one. If you had unlimited time and funding, what would you build for education in the Philippines?

Money is honestly less of a constraint than time, and not just my time, the time of all the right stakeholders to be in the room together. If they had that, my dream is that this whole exercise of demonstrating AI use cases would be run the way the challenge was run: a structured, challenge-based approach applied to the most important problems where there’s a clear case for AI. The constraint on us was that the challenge had a fixed window, so we couldn’t always get everyone in the room. 

My hope is that AI could be a leapfrog moment for Philippine education, the way mobile money was in Kenya. We never managed to put a laptop in every student’s hands, or reliable internet in every classroom. AI could let us skip some of that and put tools like virtual tutors directly in front of students and teachers.

The biggest barrier isn’t budget, because the money exists. It’s time—specifically, the time it takes to get the right people and the right part of government engaged at the right moment. That’s the brokering, really, and it can’t be rushed. Do that at scale, with the guardrails of something like the challenge, and it would be extraordinary.

Karl Satinitigan supported the Philippines team in the Ministry of Education AI Challenge, an EdTech Hub programme.

Read the next interview >

< Back to all interviews