What a strategy engagement is for
There is a point where the problem stops being “what could we do with AI” and becomes “we are doing eleven things badly across four teams and nobody can see the whole picture”.
That is what this is for. It is consulting work rather than a technology project. The output is a decision structure: what gets built, what gets bought, what gets deliberately deferred, who signs off, and what has to be true for the next phase to start.
What we look at
Process. Where the work actually flows, sourced from the people doing it rather than the process map that was drawn in 2021.
Data. What you hold, where it lives, how clean it is, and what is legally or contractually off limits. Most stalled AI programs are stuck here rather than on the model.
Systems. What your existing stack can already do, what it exposes through APIs, and where the integration cost sits. Many proposed AI builds are features you already pay for.
Capability. Who in the business can operate, extend and challenge an AI system, and what happens when that person leaves.
Risk. What a wrong output costs in each candidate workflow. This is the axis most roadmaps skip and it is the one that decides whether a workflow can run without a human in the loop.
The sequencing principle
Early items should be boring, contained and load bearing. They exist to build the muscle: the review habits, the evaluation discipline, the internal confidence that the thing works.
Ambitious items go later, once the organisation has learned to trust and to correct these systems. Roadmaps that front load the ambitious work almost always stall, because the first failure arrives before anyone has built the reflexes to handle it.
Working style
Part time and embedded. Two or three touchpoints a week with a small working group, a checkpoint with the executive sponsor every fortnight, and a live document rather than a big reveal at the end.
You will see the roadmap taking shape as it forms, and you will get to argue with it while it is still cheap to change.