Why a small East African firm can get its AI ready faster
Big firms are ahead on AI agents, but what agents need most is written context. A 20-person firm in Nairobi can produce that in weeks.
3 min read Provecta Group
Ask the owner of a 20-person firm in Nairobi how discounts get approved, and the answer is usually a name. "Ask Wanjiru." Ask how a new client is onboarded and you get a shrug and a smile: "Kamau sorts it out." Ask what the firm never promises a customer, and the owner can list five things immediately, none of which are written anywhere.
That is not a weakness for running the business. It is how most good small firms in the region work. It is, though, the single biggest obstacle to using AI agents in them.
What the size gap really is
McKinsey's 2026 State of AI survey, out in August, found that 40% of companies with more than USD 1 billion in revenue were scaling AI agents somewhere in the business, against 22% of smaller companies. The easy explanation is budget.
Budget matters. But think about what an agent needs before it can do useful work in a firm. It needs to know the clients with special terms, the current prices, who approves what and up to what amount, what the firm never says, and how it writes to a long-standing client compared with a stranger. Large firms tend to have much of that written somewhere, because at their size they were forced to: process manuals, approval matrices, compliance rules. When they build an agent, there is something to give it.
The small firm in Nairobi has the same knowledge. It lives in Wanjiru and Kamau. An agent cannot ask Wanjiru.
Where small firms have the edge
Writing that knowledge down is a very different job at 20 people than at 20,000.
The people who know are in the room. A small firm can get its owner and its two most experienced staff around a table for two afternoons. A multinational needs months of meetings across departments and countries to agree on its approval rules.
Nobody needs to sign off the sign-off. In a small firm, the owner decides the rule and it is the rule. There is no legal review of the discount policy, no regional variation to reconcile.
Change is cheap. When a price changes, one person updates one page the same day. In a large firm the same change has to travel through systems nobody fully maps.
The work is concentrated. Most small firms run on a handful of recurring tasks: quotes, follow-ups, invoices, client updates, scheduling. Write the rules for five tasks and you have covered a large share of the work.
That written set of rules and facts, kept current and loaded by every AI session before it acts, is what we call company memory. It is what Provecta Group runs its own work on, from Nairobi, and it does not need a big budget to start.
What holds firms back here
We want to be straight about the rest of the picture, because written rules are not the only gap.
Plenty of firms in the region still keep their books in spreadsheets or on paper, and their client records in WhatsApp chats and phone contacts. An agent cannot follow up on invoices that are not recorded anywhere it can read. Connectivity, the cost of AI subscriptions priced in dollars, and staff time for setup are all real constraints, and none of them is solved by writing anything down.
Data protection is another. Kenya's Data Protection Act, 2019, applies to client and staff data whether it sits in a filing cabinet or in an AI tool. Writing your firm's knowledge down is a good moment to decide what personal data belongs in it, which is usually very little.
And no one has published a measured comparison of AI results in small firms with and without written context. We rely on it in our own work, but we have not measured it, so we will not put a number on it.
Start with the names
Take the questions whose answer in your firm is a person's name. "Ask Wanjiru." "Kamau handles that." List five of them. Then sit with each of those people for half an hour and write down what they actually do, what they check, and what they would never allow. That list is the first draft of your company memory, and of your AI's instructions.
The AI readiness check helps you see which of your tasks are nearest to ready. Book 20 minutes at pgco.world/book.