Which AI projects fit the IA360 voucher (and which don't)
What the Plan IA360 business voucher literally funds, what's excluded (a bare licence) and five project types with concrete process examples.
Status as of 22 September 2026. The voucher is part of the Plan IA360, presented on 21 September 2026. There are no governing rules, no call for applications and no application window: the plan schedules the general call for the voucher, funded with €600 million, «antes de final de 2027» ("before the end of 2027"), and only after a controlled pilot planned for «primer semestre de 2027» ("in the first half of 2027"). What follows describes the criteria the document itself has already set out in writing, not a procedure you can start today.
Of the fourteen flagship projects in the Plan IA360, number 11 is the one that would directly fund a business's AI project. And the plan, for a fifteen-page document, is surprisingly specific about what kind of project it funds and what falls outside it. That line is the subject of this guide: not the procedure (which doesn't exist yet), but the criteria, which are already written down.
The criteria the voucher will use, word for word
The voucher «financiará servicios y soluciones basados en IA que sean provistos por empresas tecnológicas europeas y que utilicen los modelos fundacionales como insumo, incorporando valor añadido propio mediante desarrollo, integración, conocimiento sectorial, tratamiento de datos o rediseño de procesos» ("will fund AI-based services and solutions that are provided by European technology companies and that use foundation models as an input, incorporating their own added value through development, integration, sector knowledge, data processing or process redesign"). The key phrase is «valor añadido propio» ("their own added value"): the language or vision model is the raw material, not the product. What gets funded is the work that turns that raw material into something that solves a specific problem for a specific business.
The plan lists five ways to bring that added value. In practice they almost always overlap (few projects are only one of the five), but it's worth telling them apart because each one points to a different kind of work:
- Development: building something that didn't exist before, tailored to the specific case.
- Integration: connecting the model with the systems the business already uses.
- Sector knowledge: adapting the system to a sector's vocabulary and rules.
- Data processing: preparing, cleaning and structuring the business's own information.
- Process redesign: changing how the work gets done, not just which tool is used.
What it doesn't cover: a bare licence
The plan is just as explicit in the other direction: «el bono no financiará la mera suscripción a licencias; se dirige a servicios que transformen los procesos productivos de las empresas beneficiarias» ("the voucher will not fund the mere subscription to licences; it is aimed at services that transform beneficiary companies' production processes"). Taking out a paid assistant and handing out accounts across the workforce, without more, is exactly that: a subscription. There's no development, no integration, no redesign of anything. The plan itself doesn't say that spending is useless: it can be very useful for specific tasks; it says it isn't what this particular instrument funds.
The line isn't always obvious, and that's where the most projects get lost through being badly described: the same assistant, connected to the business's internal data and with an approval flow designed around its purchasing process, looks a lot more like what the plan calls integration with the company's own data processing than like a bare licence; the governing rules will settle that. The criterion isn't which model brand it is. It's whether someone redesigned something around it.
Five project types that fit the plan's criteria, with concrete process examples
Development with process redesign. A distribution business that receives every supplier invoice by email and types it into its ERP by hand doesn't need a generic chatbot: it needs an agent that reads the invoice, extracts the data, checks it against the original order and only asks for human confirmation when something doesn't match. The saving isn't in "having AI"; it's in that process no longer eating up two hours a day of someone who could be doing something better.
Processing your own data. A professional firm with fifteen years of case files, procedure manuals and client emails has, without realising it, an enormous document archive that no one can search faster than by asking the most senior person in the office. Turning that archive into a search system that answers by citing the exact document and page is data processing with added value: the data already existed; what gets built is the way to retrieve it.
Redesigning document processes. A business that generates invoices, delivery notes and contracts every day and files them in folders with names only the person who created them understands has a process problem, not a willpower problem. A system that classifies and routes each document to the right place the moment it arrives changes the entire filing process, rather than adding one more tool on top of the existing chaos.
Integration with sector knowledge. The same kind of system, adapted to a sector's own vocabulary and categories (an insurer's claim types, a manufacturer's product references), is where the plan's «conocimiento sectorial» ("sector knowledge") stops being an abstract phrase: it's the difference between a generic classifier and one that genuinely understands that specific business's documents.
Projects that combine several forms of added value. It's common for a project to be development, integration and redesign all at once: these aren't mutually exclusive categories, they're different angles on the same piece of work. The agent that reads the supplier invoice is development because it's custom-built, integration because it has to talk to the ERP, and process redesign because it changes who does what in the purchasing circuit. Describing a project under only one of the five categories is usually, more than a wording slip, a sign that it still isn't well defined.
The real criterion isn't the tool, it's what changes in the process
Read literally, the plan's criterion doesn't look at which model is used, but at what changes in the process: what the business did before and what it does after, and whether the difference is explained by a real change in the process or just a new subscription invoice. It's actually the same question worth asking whether or not a voucher exists: buying a tool isn't the same as building a project, and only the latter has any chance of fitting here, and also of surviving the three months after launch.
Which process is worth automating first, in a business with several candidates and limited resources, is what the test for which process to automate first ranks, without promising any savings figures.
How to prepare for this before there's a call for applications
It's worth saying plainly before we go on: no one can process an application for the IA360 voucher today, because the call for applications doesn't exist, and anyone offering to manage it or reserve you a place is selling something that isn't there. Nor can we tell you whether your project will get it or how much it would receive: that will be set by governing rules that haven't been written yet.
None of that requires waiting for the governing rules to be published: the diagnosis and choosing the use case are work you can do today, and they're the elements the plan says the rollout will include; the governing rules will set what's actually required. The full quarter-by-quarter roadmap is in your AI project's roadmap, from now to 2027 (with or without a voucher). The rest of the plan, with the other phases of the project it describes (diagnosis, training, process integration, measurement), is in what AI project the Plan IA360 will fund, and how to build it.
One more thing before we go on: we don't handle applications for the IA360 voucher, because there isn't one to handle yet, and we don't promise that a project will get it. What we can help you build is a project that meets this criterion, whether or not a call for applications ever exists.
Frequently asked questions.
Can a project covering a single small process fit, or does it need to be more ambitious?
The plan doesn't set a minimum size — it doesn't even fix the amount per business yet. The criterion it does write down is qualitative: that the provider adds value of its own and that it isn't just a licence. A project scoped to one specific, well-measured process fits that criterion better than a huge, blurry one.
What exactly does «empresas tecnológicas europeas» ("European technology companies") mean for the provider?
The plan uses that phrase without defining it any further, and as of today it doesn't set registration requirements or a provider register. It's one of the points that will depend on the governing rules, which don't exist yet.
What happens if the chosen use case doesn't work out?
The plan talks about «una primera fase que permita validar el modelo antes de su escalado» ("an initial phase to validate the model before scaling it up"), without making clear whether it means the voucher's own pilot or the pilot for each funded project. Either way, the logic of a pilot is the same: you test before committing the whole project, and that applies with or without a voucher involved.
Does automating something exactly as a person used to do it count as process redesign?
Not necessarily. The plan talks about redesign, not literally substituting a task. Automating a process exactly as it was, with the same inefficiencies, tends to end up looking more like an expensive licence than a redesign: the part that adds value is rethinking the process, not just speeding it up.
Do I need to pick the final process now, or can I start with a test one?
Nothing forces you to get it right the first time. The voucher's own design, with milestone payments and an initial validation phase the plan doesn't specify as belonging to the voucher or to each project, is built for projects that start with a narrow case and grow if it works. Starting with the process that's easiest to measure, rather than the most ambitious one, tends to produce better results than trying to cover the whole business from day one.