Your business doesn't need an AI strategy. It needs answers to five questions

The thing business leaders ask me for most is also the vaguest: help us with an AI strategy.

It usually comes after a board meeting, a conference, or a competitor announcing something. Someone senior has decided the company is behind, and everyone assumes a document will fix that. I’ve been in enough of these rooms now, with companies in fuel distribution, transport, media, interiors and education, to know how it goes. The strategy gets written. It lists technologies and quarters. It’s presented once, people say it’s good, it’s filed, and the business moves on without it.

It’s not that nobody tried hard enough. “AI strategy” just isn’t the right question. Nobody in the room is quite sure what it’s meant to answer, which is why it feels safe to pay for.

What I ask instead is five questions, in this order. I haven’t yet met a business that could answer all five and still not know what to do.

One: what do your people actually do all day? Not what the org chart says, what really happens. Where do the hours go? In most businesses I’ve seen, a lot of expensive time goes into reading something so it can be typed again somewhere else, or answering the same question for the fortieth time. You can’t point AI at your business until you can describe it at this level, and most leadership teams can’t yet. Answering this one is worth more than the strategy document.

Two: where does work sit and wait? Every business has places where nothing moves until a particular person reads something: quotes waiting to be priced, invoices waiting to be matched, complaints waiting to be sorted, reports waiting to be summarised for the boss. AI is good at that kind of work. Find those queues and you’ve found your first real uses for it, without anyone needing to name a model.

Three: if you had the answer straight away, what would change? This question kills the most projects, and it should kill them early. A dashboard that predicts demand is useless if purchasing still happens the way it always has. If the honest answer to “what would you do differently” is nothing, the AI is just decoration. Only automate decisions someone will actually act on.

Four: what do your people and customers already trust? I wrote last week about building Brik around WhatsApp and bank transfers instead of fighting them. It’s the same inside your company. A tool that sits where your staff already work gets used. A tool that asks them to move somewhere new gets dropped by February. Go for things that fit how people already work, and be wary of anything that needs your people to change first.

Five: who’s responsible for seeing it through? One person, not a committee. Every AI rollout I’ve seen work had someone whose name was on it, who used the thing every week and had the authority to change how things are done when the tool showed up a broken process. Every one I’ve seen fail belonged to everybody, so nobody really owned it. If you can’t name that person, you’re not ready, and a strategy document won’t change that.

None of these five questions mention technology. They’re about your business, and they work because AI projects fail on the business side far more often than the technical side. The models are the easy part now. Knowing where your work waits, which decisions matter and who owns them is the hard part, and a document can’t do that for you.

So when someone offers to sell you an AI strategy, ask which of these five questions it answers. And if you sit down with your leadership team and go through them honestly, you won’t need to buy one. You’ll end up with a short list of things worth doing, each tied to a queue you can point at and a person you can name.

That’s your strategy. It just doesn’t look like a document.

I write one of these every week, on AI and business in Africa as it actually works. If your company is somewhere between the board meeting and the document, it’s a conversation I have a lot, and I’m happy to have it with you.