Your AI tool works. That won’t save it.

We screened 2,300+ health tech companies down to ten for live deployment in Ghana and Malawi.

What ruled most of them out was not the product.

It was who would still pay in year three after donor left.

We had just one winner (and a close second) in the end.

A new study points at the same thing.

It mapped twenty AI tools used in African health supply chains. Of the five examined closely, only one earns revenue from its users.

The rest depend on grants.

The one paying for itself is not the biggest or most government-backed. It sells subscriptions to private pharmacies in Morocco and Spain.

Although, I’d be more calling in it a win careful because a private pharmacy is not a ministry.

What it does show is narrower: the tool that lived is the one whose buyer was already spending money on the problem.

Founders need to collect evidence of actual cost savings and not just time saved.

Without a baseline, a ministry cannot defend a permanent budget line.

So the pilot continues until the grant ends.

I used to think donor funding was a bridge to government procurement.

More often, it delays the harder question:

Who pays when the donor leaves?

Health ODA is falling. Governments are inheriting programmes they never budgeted for.

So the real question is not whether your tool works.

It is whether someone in that market was already paying to solve the same problem with something worse.

What’s that one AI-solution are you betting on scaling in Africa?