In 2020, we made a decision.
Back in 2020, we sat down and wrote our five-year strategy. One line was emphatic: we would not focus on international markets. Pakistan was enormous, complex, and more than enough to keep us busy for a decade. International expansion felt like a distraction — the kind of thing organisations pursue when they've lost focus at home.
That was the plan.
"International expansion felt like a distraction — the kind of thing organisations pursue when they've lost focus at home."
We were clear-eyed about it. The problems in Pakistan's public schools were acute, well-documented, and solvable if you committed enough time. We had committed. We were not going to get distracted.
And then the data started talking.
The data kept saying
the same thing.
Around two years ago, we went deep into our own numbers. We were looking for what separated schools that improved from ones that didn't. One pattern kept surfacing, quietly and then loudly: coaching.
Teachers who received regular coaching didn't just improve in coaching sessions. They engaged more with everything else — more lesson plans completed, more training modules used, stronger implementation across the board. The correlations were hard to argue with.
The estimated cost to scale our existing coaching model nationally. That's usually the point where funders politely look at their watches.
So we had a problem to solve. And then, almost at the same time, something else happened.
Then AI changed what was possible.
At almost exactly the same time, AI agents started arriving — capable, accessible, and genuinely useful in ways that hadn't been true even a year before. We started experimenting. One of those experiments eventually became Rumi, our AI coach.
Rumi solved the coaching cost problem.
But Rumi turned out to be the beginning of a much bigger story, not the end of one.
None of them
were pitched.
The first inbound came from Teach For Sri Lanka. They found us. We said yes and sent an early version — functional, but rough.
Around the same time, Durbeen in Pakistan ran the same beta. Their feedback was direct: this is promising, but it needs a lot more work. In hindsight, exactly what we needed to hear.
"Every organisation we talked to was facing the same wall. Coaching worked. Scaling it didn't."
Then we went to WISE in Qatar. We showed an early prototype to a handful of people. Mark from SHOFCO in Kenya was interested. So were partners working in Palestine.
A few months later, I spoke with Krupa from Silverleaf Academy in Tanzania. She understood the idea immediately — not because we explained it well, but because she recognised the problem. We applied for funding together. We got it.
The reason it resonated was simple: every organisation was facing the same problem we were. They knew coaching worked. They just couldn't afford to scale it.
Six countries.
Zero outbound pitches.
Tanzania · Palestine
active
reached
and growing
per year
Tanzania. Day one.
Krupa from Silverleaf Academy uploads Tanzania's government CPD framework — MEWAKA. Within 24 hours, Rumi has localised the platform. Here's exactly how it looked.
All plans follow MEWAKA Domain 3. I also adjusted the AI coaching observation checklist to match Tanzanian classroom norms from your notes.
Reconstructed from actual Tanzania onboarding · January 2026
The cost curve,
generation by generation.
We used to send people into schools. Fifty-five coaches across Islamabad. It worked. It cost $10 per child per year, which doesn't sound like much until you try to do it at scale.
Now we send the technology. Partners carry the last mile — their coaches, their government field officers, their existing delivery infrastructure. The cost drops to $1 to $3 per child. The evidence base stays the same. The pedagogy stays the same. The technology adapts to their context in hours, not months.
Every international partner found us. None were pitched. We're not sure what that means yet. We're watching it.
Tanzania and Palestine
launch this month.
Kenya is close. Sri Lanka is converting. Yemen is live.
The pattern we're seeing: organisations that already believe coaching works find us. They don't need convincing on the model. They need the cost to come down. That's the problem we solved.
We didn't plan to be in six countries. We planned to be in one.