Building an AI roadmap for twelve months
A roadmap is a sequence of small projects, each earning the next. How to lay out four quarters and keep the plan honest when the tools change.
The short answer
An AI roadmap for a small or mid-sized business is a sequence of small projects, laid out over four quarters, ordered so that each one earns the next: it returns measurable time or money, it builds data and access that the next project needs, and it teaches the team how to work with a system that does part of their job. The first quarter is always dull and safe. The fourth quarter is a hypothesis.
The shape of the year
| Quarter | Project type | What it must leave behind |
|---|---|---|
| Q1 | One dull, frequent, low-risk process: email triage, invoice reading, follow-ups, report assembly | A measured return, access to the systems involved, structured data as a by-product, a team that trusts the approach |
| Q2 | A second process, or the first one extended with a judgment step and a person in the loop | Logs that show accuracy, a review habit, the one-page strategy updated with what was learned |
| Q3 | The project that needed the foundations: something across systems, or with customer contact | Integrations in place, monitoring in place, evidence of value beyond time saved |
| Q4 | The project everyone wanted to start with, now cheaper and safer, or a consolidation quarter if the evidence says so | A plan for the next year written from results, not hopes |
What every project on the roadmap has
- An owner. One person accountable for it working, from the business side, not only the supplier.
- A measure. Hours per week, cases handled, days shortened, errors caught. Recorded before the project starts and again after a month in use.
- A boundary. What it does and what it does not, so success can be declared.
- A decision point. A date at which it continues, is extended, or is stopped, based on the measure.
- A person in the loop where it matters. Named, with a clear role, so the risk is bounded from day one.
Keeping it honest
- Quarterly re-plan, one hour, same people, with the measures from the last quarter on the table.
- Kill projects that did not earn. A stopped project is a saving; a zombie project is a cost and a morale problem.
- Keep the strategy page in sync. Box six of the one-page strategy is the roadmap’s next three projects; update it at each re-plan.
What this means for you
Lay out four quarters. Put the dullest high-return project in the first, choose the second so it uses what the first left behind, and let the third and fourth stay flexible. Give every project an owner, a measure and a decision date. Then re-plan every quarter with results in hand. Twelve months later you will have three or four working systems and a roadmap for the next year that is based on evidence.
Frequently asked questions
Why not plan the whole year in detail?
Because the third quarter's project will be shaped by what the first two taught you, and by tools that may not exist yet. Plan the next quarter in detail, the one after in outline, and the rest as direction. Re-plan at each quarter boundary with the evidence in hand.
How many projects per quarter?
One that matters, sometimes two if one is small. A quarter is about twelve working weeks, and a project needs time to be built, used and measured before it teaches anything. Three projects in parallel means three half-finished experiments.
What if the first project fails?
Then the roadmap's most important safeguard worked: it failed small, early and cheaply, and it taught you something about your data or your processes. Adjust quarter two accordingly. The roadmaps that go wrong are the ones where the first project was large.