Senior leadership, made concrete · Article 1 of 4
A roadmap is not a wish list of tools
Most technology roadmaps are shopping lists with dates on them. A real one is a set of decisions, tied to business goals, sequenced so that each phase earns the next.
1 · The document most firms don’t have
Ask to see a company’s technology roadmap and you generally get one of three things back.
A renewal calendar, which is a list of dates on which money leaves the building. A vendor’s slide deck with the client’s logo dropped into the corner, which is a list of things somebody would like to sell. Or a wish list assembled from conference notes and accumulated frustration: new CRM someday, look at AI, fix the file server. All three are real documents that took real effort. None is a roadmap, because none can answer the only questions that matter, which are why these things, why in this order, and what happens to the business if we stop after step two.
The absence is structural rather than embarrassing. A roadmap is a leadership artifact, and the organizations we serve, sitting between “too big for ad-hoc IT” and “too small for a full-time CTO,” have nobody whose job is to write one. So the technology plan lives across a dozen heads and a budget spreadsheet, and every year the firm makes another round of six-figure decisions on instinct and vendor charm.
A working definition before the method. A technology roadmap is a multi-year set of decisions, tied to your business goals, with costs, owners, and decision points, written so that the person who signs it can read it. Everything below serves that last clause. A plan the owner cannot interrogate is not a plan. It is a request for faith.
2 · Stage one is seeing clearly
You cannot chart a course from a position you have not fixed.
Every roadmap we build opens with a current-state diagnostic, and the diagnostic opens with people rather than servers.
Stakeholder interviews come first, in plain language, and their first job is to get the goals out of people’s heads and onto paper. What does the business want to be true in three years. Where does work drag today, and who feels it. What breaks, and who is the person who ends up fixing it at eleven at night. What does the owner actually worry about. The technology answers are hiding inside those sentences, which is why the interviews cannot be skipped and cannot be turned into a survey.
Then the systems inventory, duller and equally load-bearing: what you run, what each thing costs annually, what depends on what, where the data lives, which contracts renew when, and who holds the keys. Firms are routinely surprised by their own inventory. The subscriptions nobody canceled. The server one person understands. The two systems doing the same job for two departments that stopped talking in 2021.
What comes out of stage one is a findings report, and its first section is the alignment: where the business stands today against the goals it just named, and what the distance between the two is costing. Then the landscape as it is, the risks in business terms rather than acronyms, and the opportunities ranked by what they would do for those goals. Not recommendations about what to buy. A clear-eyed account of where you stand, written to be read by the people who have to act on it.
Note where the goals sit. They are named in stage one, because a diagnostic with nothing to measure against is only an inventory. Stage two does not discover the goals. It commits to them, with a target state, a sequence, and a number.
3 · The gate in the middle is the honest part
Between diagnosis and prescription we put a full stop, on purpose, and it is worth pausing on why, because this is the place where our incentives and yours are forced into the open.
A diagnostic that exists to justify a pre-decided phase two is not a diagnostic. It is a sales document with interviews attached, and a great deal of the industry produces exactly that. So our stage one stands alone by design. It is priced as its own engagement, its deliverables are complete in themselves, and the contract says in writing that you may conclude there and keep everything.
Some clients do. When a findings report shows a firm that its house is fundamentally sound and needs a short list of fixes any competent provider could handle, the correct next step is that short list rather than a transformation program, and we say so out loud.
The gate does something for the roadmap’s quality as well. Knowing the client can walk changes what gets written into it. Findings have to stand on evidence rather than momentum, and stage two, if it happens at all, begins from a mandate somebody actually chose rather than drifted into. Every engagement starts small and earns its next phase. The gate is where the earning gets checked.
4 · Stage two is choosing deliberately
With the landscape fixed, stage two makes the decisions, and a decision has parts that a wish does not.
A target state, tied to goals. Where the technology should be in two to three years, expressed in the business’s own terms: what a client experiences, what staff stop doing by hand, what risk gets retired, what the cost envelope looks like at the end. If a target-state item cannot be traced back to something somebody said in a stage-one interview, it does not go in. That single rule kills the wish list.
Build, buy, and retire calls with inspectable reasoning. For each significant capability the choice is to adopt a platform, build something scoped, or take the underused third option and retire what no longer earns its keep. Each call arrives with its reasoning laid out, the alternatives considered, and costs over time rather than sticker price. We hold no reseller relationships and take no commissions, so a recommendation to buy is a conclusion about your situation rather than a channel strategy. The next article has more to say about why that independence matters.
A phased sequence with owners and decision points. The roadmap lands as phases, each one small enough to fund and staff, each with named owners, indicative costs, and a defined result the owner can verify. Between the phases sit decision points, the same gate logic at smaller scale: continue, adjust, or stop, informed by what the last phase proved. And the first phase gets chosen so that it is felt soon, because a roadmap whose earliest visible result is eighteen months out will not survive its first budget season.
A written list of what you will not do. Every option the roadmap considered and declined, each with the reason it was declined. This is the section clients skip on the first read and return to most often afterwards. It earns its place on the afternoon a vendor arrives pitching something the firm already weighed, because the answer is on paper, dated, with the reasoning attached, and the meeting ends in ten minutes instead of reopening a decision that was already made.
5 · How to tell whether a roadmap is real
Whether or not we write yours, here is the inspection kit. A real roadmap passes four tests.
A new leader could pick it up and run the firm’s technology from it, which tests legibility. Every line traces to a business goal, which tests honesty. It says what you will not do, which tests judgment. And it survives contact with a vendor pitch, which tests whether it is useful at all: when something shiny arrives, the document either answers it already or gives you the frame to answer it within the hour.
One more property separates the real ones from the ceremonial. They are maintained. A roadmap is a standing decision record, revisited on a cadence, adjusted when the business changes, consulted before money moves. Wish lists get printed once and admired. Roadmaps get dog-eared.
Where this fits
This opens a short series on how the rest of our work is shaped, past the AI governance wedge: fractional technology leadership, roadmap execution, and the engagement architecture holding them together. The two-stage method above is exactly how our roadmap engagements run, fixed fee per stage, findings you keep either way.
The next article is about the judgment that writes documents like these, and what it looks like to rent that judgment two days a week instead of hiring it five.