- A realistic AI roadmap fits on one or two pages and is judged by its first executed decision, not its volume.
- Every idea passes through the same four-criterion filter: problem, value, data, effort.
- We verify data for the specific use case, not the entire company: an idea where the data isn't ready receives a "prepare" decision.
- Every initiative results in an explicit decision: launch, prepare, refine, or postpone.
- The PlanAxion AI workshop produces this roadmap in four weeks, with a decision package ready for leadership.
A realistic AI roadmap is built in four weeks, based on the operational friction your teams are already experiencing, rather than on a technological overhaul. It fits on one or two pages: use cases ranked by four criteria (problem, value, data, effort), an initial measurable pilot, and a 90-day decision sequence. Heavy transformation is not a prerequisite: it is a risk to be avoided.
A realistic AI roadmap is a short document that ranks an organization's AI use cases based on four criteria (problem, value, data, effort), designates an initial measurable pilot anchored in existing systems like an ERP, and sets a 90-day decision sequence, without requiring a prior technological overhaul.
Why do most AI roadmaps end up in a drawer?
Because they are judged on their ambition rather than on their first executed decision. A 40-page plan promising a three-year transformation doesn't survive the first executive committee meeting: no one can say what decision to make on Monday morning. Our position: a roadmap is judged by its first executed and measured decision, not by its volume.
What does a realistic AI roadmap contain?
Six elements, nothing more.
- Prioritized use cases, ranked using the same criteria.
- Documented hypotheses: expected gains and conditions for success.
- Identified dependencies: data, systems, and constraints to be addressed.
- An explicit decision for each idea: launch, prepare, reframe, or postpone.
- An initial pilot with its success metric defined before launch.
- A lead and an assignable next step for each initiative.
How do you prioritize without transforming everything?
By running every idea through the same four-question filter. Problem: is it real, specific, and experienced by an identifiable team? Value: can we measure the gain in time, quality, costs, revenue, or risk? Data: is it accessible and reliable enough to test without a prior project? Effort: is the first test proportional to the expected gain?
A weakness doesn't disappear just because the idea is appealing. The filter isn't meant to make every idea a winner: it's meant to choose which ones to launch, prepare, or postpone. It is the same filter we apply in the AI workshop.
Do you need to fix all your data before starting?
No. You verify the data required by the specific use case, not the entire company. An idea for which the data isn't ready receives a "prepare" decision, with the specific dependency to be addressed, instead of triggering a six-month project that blocks the entire portfolio. That is the difference between a roadmap and a transformation plan. In most SMEs, the cleanest data is already in the ERP: orders, invoices, payments, and stock movements. That is where the first pilots find their material.
What does the result look like? An example.
Fictional scenario, three ideas, three different decisions. Automatically categorize customer service emails: specific problem, accessible data, measurable result. Decision: launch a limited pilot. Forecast demand by product: high value but incomplete history. Decision: ensure data reliability first. Deploy an AI assistant everywhere: problem too vague, success impossible to measure. Decision: postpone and define a specific use case.
Three lines, three defensible decisions: that is a realistic roadmap.
Frequently asked questions
How long does it take to build an AI roadmap?
Four weeks is enough when the scope is limited to one business unit, with about five participants and a shared prioritization filter.
Does an SME need to hire a data team before getting started?
No. The first pilot relies on data already available in existing systems, such as an ERP. Skill requirements become clearer after the first measurable result.
What is the difference between an AI strategy and an AI roadmap?
The strategy sets the direction and ambitions. The roadmap lists concrete decisions: which use cases, in what order, with what metrics, and who is responsible.
Next step: the PlanAxion AI workshop produces this roadmap in four weeks, with a decision package ready for leadership.
