AI Strategy and Roadmap
A sequenced plan for what to build, in what order, at what cost, with the dependencies made explicit.
An AI strategy that lists twenty opportunities and calls them all high priority is a wish list. A roadmap is different. It says what gets built first, what that unlocks, what it costs to run for three years, and what happens to the plan if a key assumption turns out wrong.
What a roadmap has to answer
Before we write anything we agree the questions the document must be able to answer under pressure. Typically these are the ones a CFO or a board asks in the first ten minutes:
- Which initiative comes first, and why that one rather than the more visible option?
- What does each initiative cost to build, and separately, to run for three years?
- Which initiatives depend on data or platform work that has not been funded yet?
- What is the failure mode of each, and what is the cost of that failure?
- Which regulatory obligations attach, and at what point do they start to bite?
- What are we choosing not to do, and what does that cost us?
How we build it
Grounding
We start from your commercial objectives, not from a catalogue of AI capabilities. If the objective for the year is reducing cost to serve, initiatives that improve a metric nobody is measured on get deprioritised regardless of how technically interesting they are.
Sequencing
Initiatives are ordered by dependency and by learning value, not by expected return alone. A smaller project that establishes an evaluation practice and a deployment path is often worth more in month three than a larger one that stalls in procurement. We make that trade explicit rather than burying it.
Costing
Every initiative gets a build estimate and a three year run estimate. Run cost includes inference, storage, monitoring, retraining cycles, and the internal headcount time that most business cases quietly omit. Inference pricing is modelled at your projected volume with a sensitivity range, because unit prices move.
Governance
Each initiative is classified for regulatory exposure at the point it is proposed rather than the week before launch. Under the EU AI Act, classification changes what you are obliged to document and test, and that materially changes cost and timeline. Finding out late is expensive.
What you receive
- A sequenced roadmap with named initiatives, owners, and quarter level timing.
- A three year cost model per initiative, build and run separated, with assumptions listed and editable.
- A dependency map showing which data, platform and organisational work gates which initiative.
- A risk and regulatory classification for each initiative, with the obligations that follow.
- A one page version for the board and the full working document for the people delivering it.
The roadmap is yours to change
We hand over the cost model as a working spreadsheet with live assumptions rather than a locked PDF. When your volumes change or a model price drops, you update the inputs yourself. A roadmap you cannot recalculate is out of date the month it is delivered.
Who this suits
Organisations with real budget and competing internal proposals, where the problem is not a shortage of ideas but the absence of an agreed order. It also suits teams that need to defend an AI programme to a sceptical board, because the cost model and the explicit list of things you are choosing not to do are usually what wins those meetings.
Output
Roadmap, cost model, dependency map, governance plan. Delivered in editable formats. Full IP transfers to you on final payment.
FAQ
Do we need a readiness assessment before the roadmap?
Not always. If your data and platform position is already well understood internally, we can build the roadmap directly. If nobody can answer basic questions about data quality or integration surface, the roadmap will rest on guesses, and a two week assessment first is the cheaper path.
How far ahead should an AI roadmap look?
Eighteen months of specificity, with a directional view to three years. Anything more detailed than that in a field moving this fast is theatre. We build the document so the later horizon can be revised without rewriting the near term plan.
Will you recommend specific vendors or models in the roadmap?
We name candidate approaches and give indicative costs, but formal vendor and model selection is separate work with its own benchmarking. Committing to a specific model inside a strategy document tends to age badly.
Can you present this to our board?
Yes. Board presentation is included, and we prepare for the hostile questions rather than the friendly ones. The one page summary is written for people who will not read the appendix.
Often paired with
AI Readiness Assessment
A two week audit that scores whether your organisation can actually support the AI you want to build, and tells you what to fix first.
Read moreAI ROI Analysis and Business Case
A defensible financial model for an AI initiative, built on your numbers, with the assumptions exposed rather than hidden.
Read moreAI Use Case Discovery
A structured workshop that converts a long list of AI ideas into a ranked shortlist with feasibility and value scored.
Read moreIs this the right engagement?
Tell us what you are trying to decide. If a different service fits better, or if you do not need us at all, we will say so.