We publish positions, not commentary.

The most common way AI work fails inside a company is not technical. It is a decision, made early and defended forever, not to attach the work to a number. Once that decision is taken, the only thing left to judge the project by is activity, and there is always more activity: it never runs out.

A number is a limit, which is why it is unpopular. It rules out claiming success later on vaguer grounds. It means someone has to say, at the start, what the fix is worth, and someone else has to agree that it is worth doing at that price. That is uncomfortable in exactly the way that useful conversations are uncomfortable.

The hard part was never the model itself. A system that makes or shapes decisions needs what every other part of a business needs and rarely gets: a named owner, clear limits on what it is allowed to do, records that show how it is behaving, and a written answer to what happens when it is wrong.

Most of the failures we see were not technical. They failed because nobody could say who was answerable for what the system produced, or because the data underneath it did not hold up when someone checked it, or because the system sat beside the daily work instead of inside it.

Money can be raised. People can be hired. Knowing something others do not know wears off, usually faster than the plan that rested on it. The one advantage nobody sells is not having to care how long something takes. It cannot be bought. It can only be built into how the company is owned, and then not given away.

How long we hold a business is not a matter of taste. It follows from whose money is in it. The moment there is someone waiting to be paid back, there is a clock, and the clock quietly decides which repairs are worth making. Every fund we have watched make an obviously short-sighted decision was doing the sensible thing for the clock it was on.

Investors have spent two decades treating Africa as a single category. A category is a useful thing to raise a fund against and a useless thing to base a decision on. What makes a business something we can put money behind is never a strong opinion about a landmass. It is knowing which regulator moves slowly, which bank actually moves money on time, which two suppliers hold the whole thing up, and who picks up the phone when a shipment does not arrive.

The starting conditions did change. Mobile money, mobile identity, and mobile delivery reset what a small business needs before it can trade at all, and much of what required a branch network twenty years ago now needs a handset. That is a change in what things cost, not a story about them. It is also uneven from one country to the next, which is why the thing we choose has to be a single market and not a continent.