AI opportunity audit
We map your processes, attach real time and cost to each, and score every candidate automation on value, effort and risk. The output is a ranked shortlist, not a wish list.
Service · Strategy
AI strategy that survives contact with your business
Somebody on your board has asked what you are doing about AI. The wrong answer is to buy something. The right answer is a short, honest, costed view of where AI will make you money here — and what you should ignore.
Opportunity auditCosted roadmapROI modelBuild vs buyAI policy & governance
What is AI strategy consulting?
AI strategy consulting is the work of identifying where artificial intelligence will create measurable value in a specific organisation, quantifying that value against the cost and risk of delivering it, and sequencing the work into a plan the business can actually execute. A good AI strategy is mostly a series of decisions — what to do first, what to buy rather than build, what to leave alone — not a description of the technology.
Most AI strategy documents fail for the same reason: they describe an industry rather than a company. If a report could be handed to your competitor without changing a word, it is not a strategy, it is a magazine article.
So we start inside your operation. What does each recurring process cost you today in hours and salary? Where do errors happen and what do they cost to fix? Which decisions are late because the data arrives late? Those numbers are the entire basis of the recommendation, and they are also how you will judge us afterwards.
Signs you need this
- Your board has asked for “an AI strategy” and nobody knows where to start.
- You have three vendor proposals and no way to compare them.
- A pilot ran last year and quietly died.
- Staff are using ChatGPT with company data and there is no policy.
- You are about to spend serious money and want a second opinion first.
- You need something a board or donor can approve, in language they trust.
In practice
What a strategy engagement covers
Pick the whole programme or just the piece you are stuck on. Each of these stands alone.
Business case & ROI model
A spreadsheet you can interrogate: build cost, licence cost, running cost, internal time, expected saving, payback period and sensitivity to the assumptions we are least sure about.
12-month roadmap
What to do in each quarter, in what order, with dependencies made explicit — usually data work first, because that is what quietly blocks everything else.
Build vs buy analysis
For each opportunity: is there a credible product already, what does it cost at your scale, what does it not do, and what is the switching cost if you outgrow it?
AI governance & policy
A usable AI acceptable-use policy, a data classification scheme, an approval route for new tools, and the controls the Kenya Data Protection Act, 2019 expects you to be able to evidence.
Board & executive briefing
A 90-minute session that leaves your leadership genuinely informed rather than merely impressed — including a clear-eyed view of what AI cannot do for you.
The same answer, two ways
How we work out what is actually worth doing
We follow the money, not the technology.
First we ask a boring question: where does time go? We sit with your team and work through the week — who does what, how often, how long it takes, and what happens when it goes wrong. Most managers are surprised by the answers, because the expensive work is rarely the visible work.
Then we put shillings against it. If two people spend six hours a week each on reconciliation, that is a number. If late reports mean you reorder stock a week late, that is a number too. Now we can compare opportunities honestly instead of by which one sounds most exciting.
Next we ask what it would cost to fix each one — building it, licences, running it, and the time your own people must give. Some good-looking ideas die here, and that is the point. It is much cheaper to kill an idea in a spreadsheet than after six months of development.
What survives goes into a simple plan: do this first, then this, and here is why that order. You get the plan, the numbers behind it, and the list of things we recommend you do not do — which is usually the most valuable page.
What we will not do
We will not produce a 90-page document full of global statistics about a market you do not operate in. If the report could be handed to your competitor unchanged, we have failed.
Process-level cost accounting, feasibility scoring against data and integration reality, then expected-value sequencing under explicit uncertainty.
Baseline. Time-and-motion sampling across recurring processes: volume, cycle time, touch count, first-pass yield, rework rate, and fully-loaded cost per transaction. Where instrumentation exists we pull it (ticket systems, ERP timestamps, call logs); where it does not we sample rather than guess, and we mark the confidence interval.
Feasibility. Each candidate scored on: data availability and quality, integration surface (documented API, undocumented API, database, screen scraping, none), task determinism, error tolerance, regulatory exposure, and change-management difficulty. Determinism and error tolerance are the two that most often kill an otherwise attractive use case.
Economics. Build estimate in engineer-weeks with a range, not a point. Run-rate modelling for inference, orchestration, storage and per-conversation channel fees at your actual volumes, with headroom for growth. Internal opportunity cost included explicitly — the client-side time is the cost most business cases quietly omit and the one that most often causes a project to stall.
Prioritisation. Expected value adjusted for probability of successful delivery, then sequenced with dependencies: data plumbing and identity resolution usually gate everything downstream, so they come first even though they demo badly. We look for a fast, visible first win to fund political capital for the unglamorous work behind it.
Governance. Model and tool inventory, data classification, retention schedules, lawful basis mapping under the Data Protection Act, 2019, cross-border transfer analysis where inference happens outside Kenya, human-oversight requirements per use case, and an incident and rollback procedure.
Deliverable. A decision document with the model spreadsheet attached, the interview evidence, and a one-page summary your board can act on. Everything is yours, in editable formats.
What you get
What you walk away with
Decisions and numbers you can defend — in a format your finance director will accept.
- A ranked shortlist of AI opportunities, scored on value, effort and risk.
- A cost and ROI model you can open, edit and challenge.
- A sequenced 12-month roadmap with quarterly milestones.
- A build-versus-buy recommendation for each opportunity.
- A data and governance gap list, with the fixes prioritised.
- A draft AI acceptable-use policy for your staff.
- A one-page board summary and a 90-minute briefing to present it.
Typical engagement
Timeline1–4 weeks depending on scope
Indicative investmentFrom KES 75,000 for a focused audit
Who we work withBoards, CEOs, COOs, CIOs and finance leaders
You ownCode, prompts, docs and accounts — outright
Every quote is fixed before work starts. If we scoped it wrong, that is our risk, not yours. See how pricing works.
Typical stack
Process mappingCost modellingVendor evaluationData maturity assessmentDPA 2019 gap analysisExecutive facilitation
We choose tools your team can maintain, not the ones that make us look clever.
Questions
AI Strategy Consulting — your questions answered
What does an AI strategy actually contain?
A useful AI strategy is short and specific. Ours contains: an inventory of your processes with time and cost attached to each; a scored shortlist of AI opportunities ranked by value, effort and risk; a sequenced 12-month roadmap with what to do in each quarter; a cost model covering build, licences, running costs and the internal time you will need to commit; a build-versus-buy recommendation for each item; a data and governance gap list; and the honest list of things you should not do. It is typically 15 to 25 pages, not 90.
We already bought an AI tool and nothing changed. What went wrong?
Usually one of four things: the tool solved a problem you did not actually have; nobody changed the underlying process, so the tool became an extra step; the data it needed was not available or not clean; or nobody owned it after the vendor left. This is the most common story we hear in Kenya, and it is why we start with process and ownership rather than with product selection.
How long does an AI audit take?
One to two weeks for a focused audit of a department or a specific process area, and three to four weeks for an organisation-wide review. That includes interviews with the people doing the work, sitting with the systems, reviewing a sample of real documents and transactions, and building the cost model. We deliberately keep it short — a three-month strategy engagement is usually a sign that somebody is billing for thinking rather than deciding.
Are you independent, or do you resell software?
Independent. We take no commissions or referral fees from software vendors, and we will tell you when the right answer is a KES 3,000 per month off-the-shelf tool rather than anything we would build. If you would prefer, we will state that in writing in the engagement letter.
Do we need an AI policy?
If your staff are already using AI tools — and they are, whether or not you know it — then yes, urgently. The immediate risk is not runaway robots; it is an employee pasting a customer list, a draft contract or patient information into a public chatbot. A workable AI policy sets out what may and may not be put into which tools, who approves new tools, how outputs must be checked before use, and how this interacts with your obligations under the Kenya Data Protection Act, 2019. We can draft one in a week.
Will you also build what you recommend?
We can, and most clients ask us to — but the strategy is deliberately written so that any competent firm could execute it, and you own it outright. If you would rather take it elsewhere or build in-house, that is a perfectly good outcome and we will hand it over cleanly.
Keep exploring
Other ways we help
Most clients combine two or three of these.
AI Workflow Automation
We find the repetitive work eating your team's week — reconciliation, data entry, approvals, reporting — and hand it to software that never gets tired.
Explore AI Workflow Automation
AI Agents & Chatbots
A WhatsApp and website assistant that answers customers in English and Kiswahili, qualifies leads, checks order status and escalates to a human when it should.
Explore AI Agents & Chatbots
Custom AI Development
When off-the-shelf will not fit: document intelligence over your own files, forecasting on your own history, and integrations into the systems you already run.
Explore Custom AI Development