Consulting Desk · AI Strategy & Roadmap
Where AI creates value here,
in what order, at what cost.
And what has to be true first.
The board-ready version. Four to six quarters, sequenced by dependency, costed per phase, with the precondition named for every step. Written to be read by a board and defended in the meeting afterwards — every number has a source and every sequence has a reason.
You hold: the costed roadmap, the board paper and an investment case per phase.
What makes a roadmap defensible
Every number has a source. Every sequence has a reason.
A roadmap survives its first board meeting or it does not. The ones that fail do so in a predictable way: a director asks where a figure came from, or why phase three follows phase two, and the answer is a shrug. After that the document is decoration.
So this one is built to be interrogated. Each phase carries its cost with the basis stated. Each sequence carries the dependency that forces it — this cannot start until that data is reachable, this needs the governance route from the previous phase to exist. And each step names the precondition that has to be true before it can begin, which is what turns a roadmap into something a management team can actually run against.
We also write down the assumptions it would fall over on. Not as a disclaimer — as a monitoring list. If one of them turns out to be false, you want to find out from your own review, not from a missed phase.
It can continue straight into delivery
Where a roadmap’s first phase is clear and funded, this service runs as an Advisory Sprint that continues into a Production Build. That is a choice made at the end of the sprint, on the evidence — not a build assumed at the point of sale.
How it runs
Sequenced by dependency, costed per phase.
Built on the opportunity and readiness work rather than in place of it — a roadmap without either is a set of ambitions in date order.
- EngagementAdvisory Sprint, continuing into Production Build where phase one is clear
- MethodSequenced by dependency, costed per phase
- GateEvery step names its precondition
- Built onOpportunity and readiness findings
- Ends inA board paper, and an investment case per phase
Establish the two inputs
What is worth doing, and what this organisation can support. Where those are not already established, they are the first two weeks — a roadmap built on assumption sequences the wrong things confidently.
Find the real dependencies
Which steps genuinely block which. Most published AI roadmaps are ordered by ambition; the useful ordering is the one imposed by data, governance and skills.
Cost each phase, with a basis
What it takes to run each phase, stated so a CFO can challenge a specific line rather than the whole document.
Name the precondition per step
What has to be true before this can start. This is the line management teams actually use, because it converts a plan into a set of things someone can go and make true.
Write the board paper
The version that gets read: the argument, the sequence, the money, the risks, and the assumptions we would want monitored. Plus the alternative we considered and rejected, named.
The two lists that matter
What you hold, and what we will not do.
Both are in the engagement letter before you sign it. The second list is the one worth reading twice — it is where most disappointment in this market actually comes from.
What you hold at the end
- The costed roadmap, with dependencies made explicit
- The board paper — written to be read and defended, not translated
- An investment case per phase, with the basis for each figure
- The precondition that has to be true before each step starts
- The assumptions it would fall over on, as a monitoring list
What we won’t do
- Roadmaps ordered by ambition rather than dependency
- Numbers without a stated source
- Steps with no named precondition
- Advisory priced as the front end of a build we have already assumed
- Recommendations that only work if we are the ones delivering them
The honest exit
“Not this year” is a completed engagement.
Sometimes the defensible roadmap says wait — the data is not there, the governance route does not exist, the market has not settled. That is a legitimate output and it is contracted as one. You hold the evidence, the risk view and the reason, which is considerably cheaper than a phase one that fails for a knowable cause.
See how we contract →The build/buy question
A roadmap should say what to build and what to buy. A Build, Buy or Partner Review is the same desk, run against written supplier claims rather than a demo the supplier controlled — see the full desk.
Before you commit
Where the numbers come from, and how far out it goes.
How far out does the roadmap go?
Four to six quarters, sequenced by dependency rather than by ambition. Every step names the precondition it depends on, because a roadmap whose third step assumes a hire that has not happened is fiction.
Where do the costs come from?
Every figure has a source and the basis is stated per phase. Where a cost cannot be sourced it is given as a range with the assumption named, rather than as a single number that looks more certain than it is.
What if the answer is not this year?
That completes the engagement. “Not this year”, written down with the condition that would change it, is a defensible board position and a cheaper one than a programme nobody can fund.
Bring us the roadmap you cannot defend yet.
Thirty minutes with a founder. If you already have a roadmap, bring it — we will tell you where a board is going to push on it before they do.
Fixed fee · Quoted in writing before we start · NDA available