How to Build an AI Roadmap That Survives Contact With Reality

No plan survives first contact. Soldiers learn that early, and AI roadmaps prove it monthly.

Every mid-market company has an AI roadmap now. Most are shelf documents: thirty slides, ten initiatives, zero owners. The difference between a roadmap that ships and a roadmap that gathers dust is whether it was built to be executed or built to be presented.

Here is how we build the first kind.

Start With an Inventory, Not a Vision

Skip the vision workshop. Start by listing the workflows your company actually runs: intake, quoting, scheduling, follow-up, reporting, reconciliation. The unglamorous machinery.

For each one, capture three numbers. Hours spent per week. How often it happens. What an error costs.

Multiply. Rank. That ranked list is more strategy than most companies ever produce, and it took a spreadsheet, not a consultancy.

Pick One. Not Five. One.

Prioritize and execute. The roadmap fails the moment it becomes a portfolio of parallel pilots, because parallel pilots split your best people five ways and give every initiative a plausible excuse for stalling.

Take the top item on the ranked list that also passes two filters: the data it needs is accessible, and a named person owns the workflow today. Highest value alone is not enough. Value times feasibility wins.

Ship Inside 30 Days

Scope the first automation so it can be doing real work within a month. Not a proof of concept. Real work, in the real workflow, measured against the baseline you captured in the inventory.

Thirty days is not arbitrary. It is short enough that the organization stays interested, and long enough to do it right if the scope is honest. If the project cannot fit in thirty days, the scope is wrong, not the timeline.

Measure Against the Baseline, In Public

Whatever number the inventory said this workflow costs, report the new number every week where the whole leadership team can see it. Hours saved, response time, error rate. Pick the one that matters and publish it.

Public metrics do two things. They kill zombie projects fast, and they build the political capital you will need for initiative number two. Alignment is not a kickoff meeting. It is a scoreboard everyone can read.

Fold In Governance As You Go, Not After

Each automation ships with three artifacts or it does not ship: who can access it, what data it touches, and where its actions are logged. One page covers it.

Bolting governance on later is how companies end up freezing every AI project the first time legal asks a question nobody can answer. Build the answer into the delivery and the freeze never comes.

Then, and Only Then, Expand

Automation one is live and measured. Now return to the ranked list and take the next item. The cadence becomes a rhythm: inventory stays current, one initiative in flight, one in scoping, metrics in public.

Six months of that rhythm beats any thirty-slide vision. You will have three or four automations doing real work, a team that believes, and numbers your CFO signs off on. That is a transformation. It just does not look like the keynote version.

The Test of a Real Roadmap

Hand your roadmap to an outsider and ask them three questions. What ships in the next thirty days? Who owns it? What number does it move?

If the document cannot answer all three, it is not a roadmap. It is a brochure.

Plans are cheap. Contact with reality is where the compounding starts. Build for contact.