AI Strategy
A plan for where AI will actually help your business, and where it will not.
What is AI strategy?
An AI strategy is a written document linking AI spending to specific commercial outcomes. It sets the order of work, states expected cost and return, defines how success will be measured, and names what the business will deliberately not do.
We start from your trading position rather than the technology. Which jobs cost more than they earn. Which orders you turn away because capacity runs out. Which complaint arrives every single week.
AI earns its place only where it shifts one of those. Everything else is activity that looks like progress, and it is usually where the budget quietly goes.
What you receive is a document your managers can act on and your accountant can pull apart. It names what to build, in what order, at what price, and what has to be true for the return to appear.
Why we write strategies we can be judged on
Duara writes AI strategies for Kenyan companies from our base in Nairobi. We sit with the people who hold the budget and the people who run the operation, because a plan agreed by only one of those groups does not survive its first quarter.
We write strategies we would be willing to be judged against later. That is why every recommendation carries a number you can check afterwards, and why the document names the things we think you should refuse.
Who this is for
- You are under pressure to have an AI position
- You ran a pilot that never reached production
- Several departments want funding from one budget
- Your board wants a plan before releasing money
Industries we serve
- Banking and financial services
- Insurance
- Telecommunications
- Manufacturing and distribution
- Education
- Government and NGOs
Problems this solves
Strategy work usually starts because one of these has already happened, and the cost of the next mistake is now visible.
- Activity without a stated goal. A competitor announces something, the board asks what you are doing, and work begins before anyone agrees what it is meant to achieve.
- Pilots that stall before production. A demonstration impresses everyone and then quietly dies, because nobody defined what moving it into daily use would require.
- Departments competing for one budget. Three proposals arrive with no shared basis for comparing them, so the loudest sponsor usually wins.
- No agreed measure of success. Nothing was recorded beforehand, so twelve months later nobody can say whether the money achieved anything.
What we deliver
A strategy document that ties every proposed piece of work to a commercial outcome, with sequence, cost, agreed success measures and an explicit list of what to avoid.
A written strategy linked to specific commercial outcomes
Opportunities ranked by what each one saves or earns
An explicit list of what we advise against pursuing
Success measures agreed before any work begins
Build cost and running cost for every recommendation
A short board paper summarising the decision required
Features
- Commercial framing before technical framing. Every recommendation attaches to a margin, a capacity ceiling or a complaint you already receive.
- A written refusal list. We name what not to do. Clients often tell us this is the section that saved them the most money.
- Justified sequencing. The order of work carries its reasoning, so changing priority later is a decision rather than a drift.
- Measures agreed in advance. What gets measured, and against which baseline, is settled before the first build rather than argued afterwards.
- Running costs modelled at real volume. Usage-priced AI costs more when it succeeds. The numbers reflect busy months, not pilot months.
Benefits
- One direction instead of several. Departments stop advancing separate proposals that quietly contradict each other.
- Board decisions get faster. Directors approve or reject against stated assumptions rather than debating in the abstract.
- Early work funds later work. Sequencing by return means the programme pays for its own next stage.
- The result is provable. Because measures were fixed upfront, the outcome is checkable rather than a matter of opinion.
Technologies we use
- Your financial and operational records. The strategy is built from your figures, not from industry averages that may not resemble your business.
- Opportunity scoring. A consistent comparison method so the ranking survives scrutiny from whoever did not get funded.
- Usage-based cost modelling. Projecting the monthly bill at genuine volume, which is where most AI budgets are underestimated.
- Benchmark review. What comparable operations achieve, used as a sanity check rather than as a promise.
- Roadmap handover format. The strategy is written to feed directly into a sequenced delivery plan without rewriting.
Strategy work is analysis rather than engineering. The only tools involved are the ones needed to interrogate your own numbers.
How the process works, step by step
Strategy work runs over three to four weeks and involves two or three sessions with your leadership team.
-
Step 1
Free positioning call
We establish what is driving the question, who holds the budget, and what a decision would need to look like.
-
Step 2
Commercial review
We go through your margins, capacity limits and recurring complaints to find where AI could genuinely move a number.
-
Step 3
Leadership workshops
Two or three working sessions to test assumptions with the people who will have to live with the plan.
-
Step 4
Drafting and costing
We write the strategy, cost each recommendation at real volume, and build the refusal list.
-
Step 5
Board presentation
We present it and take challenges directly from your directors and finance team.
Quality assurance and how we work
The commitments we make specifically on strategy work.
- Every recommendation carries a number. If we cannot say what a piece of work is expected to change, it does not go in the document.
- We write the refusal list first. Deciding what not to do is done before deciding what to do, so the plan is not padded.
- You own the document. It is yours to take to any supplier, including our competitors, without restriction.
- We declare our own interest. Where we would be a candidate to build something we recommend, the document says so on that page.
- Assumptions are separated from facts. Anything estimated is labelled, so nobody mistakes a projection for a measurement.
Frequently asked questions about AI strategy
What is an AI strategy?
A written plan linking AI investment to specific commercial outcomes, with the order of work, expected costs, agreed success measures and an explicit list of what not to do.
How does it differ from a readiness assessment?
A readiness assessment establishes what is possible with your current data and systems. A strategy decides what is worth doing and in what order.
We already have a strategy document. Will you review it?
Yes, and reviewing is quicker and cheaper than starting again. The usual gap is that nothing in it connects to a figure the business already tracks.
How long does a strategy take to produce?
Three to four weeks, including two or three working sessions with your leadership team.
Must we use Duara to deliver it?
No. The document is written to be delivered by your own team or any other supplier, and it says so.
Who needs to be in the room?
Whoever controls the budget and whoever runs the affected operation. If either is missing the plan tends not to survive.
What if the conclusion is that AI is not the answer?
Then the strategy says that. Plenty of problems are solved far more cheaply by fixing a process or buying ordinary software.
How far ahead should a strategy look?
Twelve to eighteen months. Past that, both the technology and your own priorities move enough to make the detail fictional.
Are costs included?
Yes, build cost and running cost for each recommendation, modelled at realistic volume rather than pilot volume.
How do you decide what to exclude?
Anything where the expected return fails to clear the full cost, or where the downside risk outweighs the upside.
Can a strategy cover several departments?
Yes, and it usually should, because the strongest opportunities tend to sit in the gaps between departments.
What happens when priorities change?
The strategy carries decision points. Changing direction at one is ordinary; discovering the problem too late is what costs money.
Will it name specific suppliers or products?
Where it matters, with reasoning shown. We also state on the page when we would be a candidate for that work.
Is this worthwhile for a small business?
Yes, though the document is shorter. A small business usually needs one clear priority rather than a portfolio of them.
Do you run workshops as part of it?
Yes, normally two or three sessions to test assumptions with your team before anything is written.
How do you handle internal disagreement?
We write down the competing positions and the evidence for each, so the decision gets made openly rather than by seniority.
Does it look at what competitors are doing?
As a sanity check only. Copying a competitor without knowing their economics is not a strategy, it is a guess.
What form does the deliverable take?
A written document plus a short board paper. No jargon, and no slides that say nothing.
Can you help implement afterwards?
Yes, and we declare that interest inside the document itself so the recommendation can be judged accordingly.
What is the most common mistake you see?
Beginning with the most technically interesting project, so nothing visible exists by the time the budget is next reviewed.
Glossary of terms
The vocabulary used in strategy documents, without the consulting jargon.
- AI strategy
- A written plan connecting AI investment to commercial outcomes, including what the business will not do.
- Commercial outcome
- A measurable business result such as lower cost per case, higher conversion, or capacity to take more work.
- Pilot
- A small trial used to test whether an approach works before committing to full deployment.
- Production
- A system in genuine daily use by staff or customers, as distinct from a demonstration.
- Usage-based pricing
- Charging per unit of use, so the bill rises as adoption grows. Standard for most AI services.
- Refusal list
- The explicit set of proposals the strategy recommends declining, with the reasoning attached.
Related services in AI Consulting
Talk to us about AI strategy
The positioning call costs nothing. If your question turns out to be simpler than a full strategy, we will tell you that on the call.
Start your project