17% of Basque companies use AI — and they're earning 8.7% more: what they're doing differently

While 95% of AI pilots fail globally, the Basque Country shows a model that actually works

The contrast nobody’s discussing

In my previous article about the big copilot lie, I shared devastating data: 95% of generative AI pilots show no measurable P&L impact, according to MIT. Developers using AI tools take 19% longer to complete tasks.

But there’s one data point that contradicts that pessimistic narrative — and it comes from the Basque Country.

According to the second Diagnosis of Artificial Intelligence in Euskadi, published by BAIC (Basque Artificial Intelligence Center) in October 2025:

17.4% of Basque organizations already use AI (vs. 12.2% the previous year)
User companies attribute an 8.7% average revenue increase to AI use
Adoption grew 42% in just one year
AI provider companies doubled their revenue: from 105 to 223 million euros

What is Euskadi doing differently?

The ecosystem, not the tool

The first difference is approach. While many companies adopt AI by buying Copilot or ChatGPT Enterprise licenses and hoping for magic, the Basque model is based on building an ecosystem.

BAIC isn’t just a research center. It’s a public-private alliance that includes:

Large industrial companies (like Fagor Arrasate)
Startups
ICT providers
Technology centers (Ikerlan, Tecnalia, Vicomtech)
Basque Government

The key lies in a phrase the Government keeps repeating: “Data sovereignty will be the key to our economic future.”

It’s not about using others’ AI. It’s about generating, managing, and leveraging intelligence locally.

Real investment, not just licenses

The investment numbers in Euskadi are significant:

557 million euros in total AI investment
92.6% of agents plan to continue investing
9,500 specialized professionals and 1,100 annual AI graduates
The Basque Government allocates 1.4 billion euros to Euskadi’s digital transformation

But what’s interesting isn’t just how much they invest, but in what.

SPRI’s (the Basque Country’s business development agency) “Artificial Intelligence 2026” grant program offers up to €100,000 per company, but with specific conditions:

Effective adoption projects — not exploratory pilots
Strategic data management — before implementing AI
AI Act and Data Act compliance — governance by design
Alignment with standards and certifications — responsible adoption

In other words: they don’t fund experiments. They fund real implementations with governance included.

The industrial factor

Euskadi has a structural advantage: its industrial fabric.

Unlike more service-oriented economies, the Basque Country has a strong manufacturing base — machine tools, automotive, energy, aerospace. These sectors have:

Abundant data from production processes
Clear use cases for optimization, predictive maintenance, quality control
Measurable ROI — if you reduce defects or line stoppages, the impact shows in the balance sheet

Amaia Iñurria, from Fagor Arrasate, explained at the ETEKIN meeting: companies must create “experimentation and learning spaces,” but the goal isn’t experimenting for its own sake. It’s internalizing new ways of working.

The 80% that changed their culture

One of the most revealing findings from the BAIC diagnosis is this:

80% of agents have experienced cultural change due to AI, and 20% consider it “very significant.”

This contrasts with what we see in many companies: tool adoption without process change. You buy Microsoft Copilot, install it, and expect employees to become more productive.

The Basque model is different: technology comes after organizational change, not before.

The Euskorpus project: linguistic and technological sovereignty

A concrete example of how Euskadi thinks differently is the Euskorpus project.

The Basque Government has allocated 10.55 million euros (2025-2027) to develop a quality digital corpus in Euskera that enables integrating the Basque language into AI technologies.

Why does this matter?

Because if you depend on models trained in English and Spanish, your ability to apply AI to local contexts — public administration, education, citizen services — becomes limited.

It’s applied technological sovereignty: not just using AI, but having the capacity to adapt it to your reality.

Lessons for the rest of Spain (and Latin America)

The Basque model isn’t perfectly replicable — Euskadi has an industrial history, a system of economic agreements, and a public-private collaboration culture that doesn’t exist everywhere.

But there are universal lessons:

1. Ecosystem before tools

AI doesn’t work in a vacuum. You need quality data, trained talent, and clear use cases. Investing in the ecosystem yields more returns than buying licenses.

2. Governance by design

Companies that implement AI without thinking about data governance end up with pilots that don’t scale. The European AI Act requires this, but Basque companies are doing it proactively.

3. Measure what matters

The 8.7% revenue increase is a real metric, not self-reported “time saved.” If you can’t connect AI to business results, you’re probably not measuring correctly.

4. Continuous training

9,500 specialized professionals and 1,100 annual graduates don’t appear by themselves. They require sustained investment in training — and alignment between universities, technology centers, and companies.

5. Think long-term

The Digital Transformation Strategy 2030 and Euskadi’s AI Plan aren’t shelf documents. They’re roadmaps updated with real data — like the BAIC diagnosis.

My perspective

As a data engineer working in Spain, the Basque case seems both hopeful and frustrating.

Hopeful because it demonstrates that AI can generate real value when implemented strategically.

Frustrating because in many companies we keep seeing the opposite pattern: buying tools, launching pilots without clean data, measuring with vanity metrics, then wondering why it doesn’t work.

The 17.4% adoption rate in Euskadi might seem low compared to media hype. But it’s real adoption, with measurable impact, in a sustainable ecosystem.

That’s worth more than the 78% “adoption” reported in other surveys where “using ChatGPT to write emails” counts as AI implementation.

Conclusion

The contrast between global data (95% of pilots fail) and Basque data (8.7% more revenue) isn’t coincidence.

It’s the result of:

Thinking in ecosystems, not tools
Investing in data and talent before licenses
Measuring real results, not activity
Effective public-private collaboration

The question for the rest of us is: are we willing to do the hard work this requires, or do we prefer to keep buying magic solutions?

If you want to see more examples of companies generating real ROI with AI agents, check out my analysis of AI agents for enterprises. And for the global context of where all this is heading, read the AI trends for 2026.

Sources

BAIC (October 2025). “Second Diagnosis of Artificial Intelligence in Euskadi”
Basque Government / SPRI (2025). Artificial Intelligence 2026 Grant Program
Irekia (November 2025). “Artificial Intelligence centers the debate at the last ETEKIN meeting of 2025”
Basque Government (July 2025). Euskorpus Project