Your AI project's roadmap, from now to 2027 (with or without a voucher)

Status as of 22 September 2026. The Plan IA360 was announced on 21 September. There is no call for applications for the business voucher, no governing rules and no application window: the plan places a controlled pilot in the first half of 2027 and the general call before the end of that year. Nothing that follows is a procedure. It's the roadmap for your own project, with those dates as a reference point, not a calendar to meet.

There's still nowhere to apply for the voucher. What it does have, in writing, is the method it announces: a prior maturity diagnosis, identifying a specific use case, training staff, incorporating AI into internal processes, and subsequently measuring the impact on productivity. That method doesn't need any governing rules before you can start applying it today, and it's the same thing an AI project needs even if it never sees a euro of public money: the difference between a project that survives and one that gets abandoned usually isn't whether there was funding, but whether someone did these five steps in order. This guide isn't the plan's calendar (that's in the official Plan IA360 document): it's the calendar for your project, with the plan's milestones as reference points.

Your project, quarter by quarter

Before you spend anything: what you execute and pay for before governing rules exist may not be eligible for funding. Many grants only cover what's done after you apply for them or after they're awarded. Since the IA360 governing rules don't exist, no one can tell you today whether it will count.

If your business would rather not wait and run the project with its own resources, it can bring the pilot and the rollout forward: the work is worth exactly the same with a voucher as without one. What doesn't make sense is doing it while assuming the voucher will reimburse it afterwards.

There's no obligation to follow this exact pace (every business has its own internal calendar), but lining up your own phases with the milestones the plan has already set makes sense: if the call for applications arrives as planned, your own project will already have gone through diagnosis, pilot and evaluation before any form exists to fill in. And if it's delayed relative to what the plan anticipates today (which can't be ruled out, since there are no governing rules yet), the work done isn't wasted: an evaluated use case and a measured baseline stay useful whenever it does arrive, whether earlier or later than expected.

What data needs to be ready before the diagnosis

A maturity diagnosis doesn't start by choosing a model: it starts by looking at the data and what state it's in. Reviewing that before spending a euro is precisely the first block of the maturity diagnosis and 18-month roadmap: with that map done, the diagnosis has something to work with on which processes genuinely benefit, which team will maintain them and, sometimes, whether the honest answer is that AI doesn't solve that particular problem; reorganising a process or training a person does. Saying that in time, before building anything, is what separates a useful diagnosis from one that just confirms what the client wanted to hear.

Choosing the use case: which process first

Almost no business has just one candidate: there are usually several processes that look automatable and limited resources to tackle them all at once. The common mistake is starting with the process most visible to leadership — the one that comes up most in meetings — instead of the one that best meets the technical conditions: enough volume to make it worthwhile, reasonably clear rules rather than pure expert judgement, accessible data, and enough tolerance for error to allow a human review phase while it's being calibrated. Prioritising by those criteria, rather than by which one sounds best in a presentation, avoids building the wrong pilot. That's what the test for which process to automate first ranks, without promising savings figures nobody can calculate yet.

Putting a number on the decision

The last element the plan lists is the one most often overlooked: subsequent measurement of the impact on productivity. Measuring afterwards requires having recorded beforehand how much the process cost, and that has to be done while the process is still manual, not once it has already changed. Recording the monthly hours it consumes today, the cost of that hour, and an honest estimate of the automatable percentage lets you calculate, with the automation ROI calculator, the break-even point and a pessimistic and optimistic range using your own numbers, with no preloaded figure and no invented "typical saving".

The resource that's genuinely worth setting aside now is the time of the team that will take part in the diagnosis and the pilot: that's what decides whether this roadmap actually happens or stays an intention. Money is different: the amount per business, the eligibility criteria and what a provider has to meet to fit the phrase «empresas tecnológicas europeas» ("European technology companies") are pieces the plan hasn't fixed yet, so committing budget against a voucher without those conditions written down is betting blind. None of that changes the preparation work described above: the diagnosis, the use case, the baseline and the training are worth doing regardless of how those details get settled.

One clarification worth spelling out: we don't process applications for the IA360 voucher, because there's no application to handle yet, and we won't promise you'll get it. What we do is the diagnosis, the pilot or the training in this roadmap, whether the voucher call is ever published or not.

Related pages within the same plan are which AI projects fit the IA360 voucher (and which don't) and testing AI with your own data before deploying it. The other phases of the project the full plan describes are in what AI project the Plan IA360 will fund, and how to build it.