Digital and AI Transformation Strategy

The answer nobody selling you something can give you

Boards arrive at this question in a recognisable way. Change has become expensive, so every proposal comes back priced against a system rather than against the idea. Budget requests arrive from four directions at once, each of them plausible, with no honest way to rank them against each other.

Almost everyone with an opinion has something riding on it. Your core provider's recommendation involves more core. A channel vendor's recommendation is a channel. A global advisory report describes a market unlike this one and prices the work in a currency you do not earn in. Inside the institution the strongest case wins rather than the best one, because nobody is placed to compare a branch proposal with a data proposal on the same terms.

This matters most where a wrong turn stays expensive for years. Banks facing a core renewal or a supervisory expectation they cannot meet with what they run today. SACCOs whose membership has outgrown the way it is administered. Microfinance institutions whose field operation runs on paper while the competition lends from a phone. Telcos and fintechs, where the constraint is not money but one team who can build only one of the five things being asked of them.

What an honest assessment tells you

It starts with where you actually are, which is not always where the organisation chart or last year's strategy paper says. What you can offer a customer today, and what it costs you to change any of it. Where you depend on a single supplier, and what that dependency is priced at across everything else they sell. What your own information would support if a board member or a supervisor asked a hard question of it. How much change your people can absorb alongside the jobs they already have.

What comes back is usually duller than what the board expected to discuss. A monthly figure assembled by hand from several places, so nobody can answer a simple question mid-month. A system nobody will retire because one report depends on it. A contract signed years ago that has put a price on every change since. None of that is a strategy problem. It is what makes strategy expensive, and why ambitious programmes stall in their second year.

The value is a shared and honest starting point. Institutions rarely disagree about where they want to be. They disagree about where they are, and most arguments about sequence are that argument in disguise. There is a free readiness assessment on this site covering the same ground at a high level. It is worth doing before you commission anything, including from us.

AI, with the pitch removed

AI is oversold, and financial institutions in this region are a favourite audience for it. The pattern repeats: a demonstration on somebody else's clean data, a board asking why we do not have that, a pilot funded, a room impressed, and nothing that reaches a customer. The pilot is not renewed, and the institution decides the technology does not work when what failed was the order.

The uses that pay are narrower than the pitch. They sit on work that is repetitive, high in volume and already costing you salaries, and they depend on information you would defend in writing. An institution that cannot produce a reliable count of its own customers is not ready to make automatic decisions about them. That is the unglamorous half of most transformation plans, and usually the honest answer to when to start on AI: after work you would rather not put in a board paper.

There are local realities as well. Customers deal with you in Kiswahili and English, often inside one sentence. A supervised institution has to explain a decision to the customer it affected and to its supervisor, and personal information used in reaching it falls under Tanzania's Personal Data Protection Act. The question is not whether any of this works, but which uses you could defend, staff and explain, and we will say which of the things being sold to you fall outside that line.

Options with the price and the risk attached

The output is a document, not a system. It sets out where you are, the options open to you, and a sequence. Each option carries what it costs, what it puts at risk, what has to be true for it to work, and what it closes off later. An option presented without those four things is a sales pitch with a diagram on it.

The options include the ones vendors do not present. Doing less. Renegotiating rather than replacing. Stopping a programme that is already funded and half delivered, which is the hardest recommendation to make and occasionally the right one.

Then the sequence, which matters more than the list. Work is ordered so each step either pays for the next or takes a risk out of it, and nothing done early is thrown away by something done later. It is written to be read by the people who will fund it: what it costs, what it returns, what it depends on, and what happens if you do nothing. The document is yours whether or not we do any of the work in it.

What this asks of you

A sponsor with real authority, first. Not a coordinator, but someone who can commit the institution, carry a finding through committee, and stop something already funded. Assessments commissioned below that level produce a document that circulates and changes nothing.

Then access, to people as much as to systems. The people who know how the institution really works are seldom the ones the chart nominates. The branch manager who can say which promises the network cannot keep. The person who holds a supplier relationship and remembers what was actually agreed. Their time has to be released by someone senior or it will not happen. We also need to see what you run and the contracts behind it, and to be told plainly where you are not permitted to show us something.

Last, a willingness to hear an uncomfortable answer. Sometimes the finding is that a funded project should stop, that the problem is organisational rather than technical, or that the AI programme the board has been promised should wait a year. That is the most valuable thing an assessment produces and the hardest to receive. Agree before it starts that it reports what it finds, and to whom, rather than settling that while a draft is being argued over.

Our own interest, and what we do about it

We build software, so we have an obvious interest in a plan that involves building software, and you should read anything we tell you with that in mind. Two things follow. Every set of options includes the ones where we build nothing, costed and argued as seriously as the rest. And the plan is a document you own outright, with no obligation to appoint us for any part of it.

If you go on to build with us, you see working software in an environment you control rather than reports about it, and the source code and documentation are yours from the first commit. We support what we build and train your people to take it on, and there is no lock-in. If you appoint someone else, the plan still holds: it was written for your institution, not for our proposal.