Only 4.1% of Spanish industrial SMEs use AI: what's missing

The IndesIA 2026 barometer, published this week, confirms something we see constantly in conversations with industrial SMEs at Evicron, an AI and custom software studio based in Barcelona with over 200 projects delivered across 12 sectors since 2019: only 4.1% of Spanish industrial SMEs actually run AI in their processes today, up from 2.9% in 2025 and 2.1% in 2024. The number is growing, but it’s still marginal nearly three years into generative AI being available to any business. The useful question isn’t “why so few”, it’s “what’s different about that 4.1%”.

What the barometer actually shows

According to La Ecuación Digital and Forbes España, adoption scales sharply with company size: medium-sized firms (50-250 employees) reach 6.5%, small ones (10-50 employees) 5.1%, and micro businesses (under 10) sit at just 3.6%. The barometer also flags a technology shift: generative AI and decision-flow automation are gaining ground over the more classic machine-learning approaches that used to dominate industrial settings like predictive maintenance and quality control.

For context: other barometers using a looser definition of “using AI” — one that counts occasionally running ChatGPT for marketing or customer service — put overall Spanish business adoption above 50%. These aren’t contradictory figures; they’re measuring different things. IndesIA asks about AI embedded in real production processes, not casual use of a consumer tool. That gap is precisely why so many SMEs get stuck at “we tried ChatGPT once” without ever reaching something that moves an actual business metric.

Why most companies stall at the pilot stage

Across the applied-AI projects we run at Evicron, the same pattern shows up almost every time: an industrial SME tries a generic tool — a chatbot, an office assistant — finds it useful for scattered tasks, and stops there. What’s missing isn’t willingness or budget — the EU AI Act and public funding schemes have been pushing both for two years now — it’s a specific, measurable use case tied to a process that already produces data: a production line, a warehouse, a quality department. Without that anchor, AI stays a nice-to-have office tool instead of becoming an operational lever.

What sets the 4.1% apart

Three traits keep showing up in the industrial SMEs that turn an AI pilot into something that sticks:

• They start with a process that already produces data. Machinery maintenance, quality control, stock management or shift planning make better starting points than “we want AI” in the abstract, because there’s already a track record to measure improvement against.
• They set a metric before starting. Downtime avoided, defective parts caught earlier, order response time. Without a baseline number, there’s no way to tell whether the pilot worked or just produced a nice-looking demo.
• They treat the pilot as a project with an end date, not an open-ended experiment. A discovery sprint, a clickable demo within a few weeks, and a clear go/no-go decision — not an open contract with a big consultancy billing by the hour with no commitment to a result.

How to start without a corporate budget

Closing the gap between the 4.1% and the rest doesn’t require a multinational’s budget. At Evicron we start with a free discovery session, followed by a fixed quote and a 4-phase process with weekly sprints and a clickable demo as early as week 2 — no waiting months to find out whether the idea works. Custom applied AI starts at €6,000, well below the real cost of a failed pilot repeated two or three times with different vendors, which is what actually gets expensive.

If your company works in industrial manufacturing and wants to move from “we tried some AI” to a process that actually shows up in the numbers, our AI consulting service starts exactly there: identifying the process with the most data already available and the shortest possible pilot to test it. And if the pilot is already clear and what’s left is building it, applied AI is the next step.

Want to know if your company already has the right process for a first AI pilot worth running? Get in touch: the first session is free and we reply within 24 hours.