AI adoption in Spanish companies: Spain vs EU and how to start
In 2025, 20.27% of Spanish firms use AI, slightly above the EU‑27 (19.95%), per Eurostat. See the most used technologies, size gaps and a 90‑day AI plan.
Francesc Beumala · Founder, partner and CEO of ARTIC
In this article
- Where Spain stands in 2025 vs the EU and peer countries
- The size gap: who uses more AI and why it matters
- Which AI technologies enterprises use most
- The trajectory in Spain: from 9.18% to 20.27% in two years
- Priority AI use cases and how to choose them well
- From diagnosis to first results: a 90‑day delivery plan
- AI agents and practical enterprise architecture
- How to measure AI ROI: metrics that matter
- People, training and governance: the differentiator
- Frequently asked questions
- Sources and methodology
In 2025, 20.27% of Spanish companies (10+ employees, excluding finance) used at least one AI technology, slightly above the EU‑27 average of 19.95%, according to Eurostat’s ICT usage in enterprises survey (dataset isoc_eb_ai, updated 2026‑06‑15). Large firms lead AI use in Spain at 57.46% versus 55.03% in the EU‑27. The most widespread technologies are language generation, text analysis, AI‑based automation and machine learning.
Key takeaways
- Spain reaches 20.27% of enterprises using AI in 2025, marginally above the EU‑27’s 19.95% (Eurostat, isoc_eb_ai, updated 2026‑06‑15).
- By size in Spain: 17.17% (10–49), 30.1% (50–249) and 57.46% (250+). EU‑27 is similar: 17%, 30.36% and 55.03%.
- Language generation stands at 8.92% of Spanish firms; text mining at 10.05%; AI‑based automation at 6.68%; machine learning at 5.62% (2025).
- The Netherlands (33.21%) and Germany (25.97%) outpace Spain; France (18.16%), Italy (16.4%) and Portugal (11.54%) trail.
- Spain more than doubled overall AI use from 9.18% (2023) to 20.27% (2025), with the fastest growth in language and text use cases.
Where Spain stands in 2025 vs the EU and peer countries
AI adoption in Spanish enterprises has stepped up. According to Eurostat’s survey on ICT usage and e‑commerce in enterprises (companies with 10+ employees, excluding the financial sector; dataset isoc_eb_ai, updated 2026‑06‑15), 20.27% of Spanish firms used at least one AI technology in 2025, slightly above the EU‑27 average of 19.95%.
Among reference countries, the Netherlands (33.21%) and Germany (25.97%) are well ahead of the EU average. France (18.16%) and Italy (16.4%) sit below Spain, and Portugal (11.54%) lags behind. This competitive picture shows Europe is moving and Spain has accelerated convergence in the last two years.
What does “using AI” mean here? The composite indicator includes firms using at least one of the following: technologies that generate written or spoken language or programming code; text mining; AI‑based automation to assist workflows or decision‑making; machine learning for data analysis; and other AI capabilities covered in the Eurostat indicator.
In practical terms, this level of adoption means AI is no longer confined to innovation pilots: it is entering day‑to‑day processes in marketing, customer service, operations and internal support.
For a B2B services SME in Madrid (50–120 employees), this context suggests two priorities: 1) identify quick wins where AI can save time within weeks (e.g., automatic ticket categorisation and proposal summarisation), and 2) plan deeper capabilities (AI agents connected to internal data) to strengthen an edge against European competitors.
- Netherlands
- Germany
- Spain
- EU-27
- France
- Italy
- Portugal
Source: Eurostat (isoc_eb_ai), 2025
AI use in enterprises (Eurostat, 2025)
| Country or segment | Value |
|---|---|
| Spain · Total (10 or more) | 20.27% (2025) |
| Spain · 10 to 49 employees | 17.17% (2025) |
| Spain · 50 to 249 employees | 30.1% (2025) |
| Spain · 250 or more | 57.46% (2025) |
| EU‑27 · Total (10 or more) | 19.95% (2025) |
| Germany · Total (10 or more) | 25.97% (2025) |
| France · Total (10 or more) | 18.16% (2025) |
| Italy · Total (10 or more) | 16.4% (2025) |
| Portugal · Total (10 or more) | 11.54% (2025) |
| Netherlands · Total (10 or more) | 33.21% (2025) |
The size gap: who uses more AI and why it matters
Company size remains the main predictor of adoption. In Spain, 57.46% of firms with 250 or more employees use AI, compared with 30.1% among those with 50–249 and 17.17% among 10–49. In the EU‑27 the pattern is almost identical: 55.03%, 30.36% and 17% respectively (Eurostat, isoc_eb_ai, 2025).
Why do large enterprises use more AI? Three recurring reasons: 1) more data and standardised processes; 2) larger budgets and in‑house skills to integrate AI into corporate systems; and 3) stronger regulatory and competitive pressure that drives automation and advanced analytics.
For a mid‑sized industrial firm in Valencia (around 120 employees), the takeaway is clear: competing with giants means focusing AI where marginal return is highest. Practical moves include deploying AI‑assisted automation in non‑conformity handling and work‑order preparation; using text mining on incident logs to spot common causes; and applying language generation to speed up technical documentation and sales communications.
Operational tip: prioritise domains with “data ready” (tickets, emails, work reports, orders) and run 6–8 week sprints per use case. Define input facts (volume, data quality, current time/errors) and a target output (time saved, SLA, accuracy) you can verify at the sprint’s end. If you need delivery support, an experienced partner such as ARTIC can help you frame objectives and measure results in production rather than in demos.
SpainEU-27
- 10–49 staff
- 50–249 staff
- 250+ staff
- All (10+)
Source: Eurostat (isoc_eb_ai), 2025
You can also explore delivery models with a specialised AI agency for businesses to co‑own outcomes with your internal team.
Which AI technologies enterprises use most
Eurostat’s breakdown shows several technology categories. In 2025, in Spain the stand‑outs are:
- Language or code generation and speech synthesis: 8.92% of firms (EU‑27: 8.76%).
- Analysis of written language (text mining): 10.05% (EU‑27: 11.75%).
- AI‑assisted workflow automation/decision support (AI‑based RPA): 6.68% (EU‑27: 5.35%).
- Machine learning for data analysis: 5.62% (EU‑27: 5.1%).
In large Spanish companies (250+ employees), penetration is markedly higher: 32.14% for language generation, 34.26% for text mining, 28.46% for AI‑based automation and 29.34% for machine learning (2025). Beyond budgets and talent, this difference signals that data maturity (quality, governance, access) is the true accelerator.
Examples by function:
- Marketing and sales: generate proposal drafts and follow‑ups; analyse feedback forms and reviews; score leads with ML.
- Operations: extract data from documents; prioritise incidents; detect patterns in maintenance logs.
- Customer support: assistants with speech synthesis and NLG for initial replies; text mining for reasons of contact.
- Internal support: automate repetitive admin (onboarding/offboarding, case files) and apply analytics for demand prediction.
Implementation tip: start with cases involving structured or semi‑structured text, where NLG/NLU and AI‑based RPA deliver quick returns without large data‑platform projects. If you already standardise collaboration and security around advanced models, consider a secure rollout with Claude for business.
The trajectory in Spain: from 9.18% to 20.27% in two years
Spain’s AI adoption is accelerating. The share of firms using at least one AI technology rose from 9.18% (2023) to 11.31% (2024) and 20.27% (2025). This growth tracks the diffusion of foundational technologies and the normalisation of AI in everyday processes (Eurostat, isoc_eb_ai; update 2026‑06‑15).
By technology, the pattern is similar:
- Language/code generation and speech synthesis: 2.29% (2023) → 4.27% (2024) → 8.92% (2025).
- Text mining: 3.07% (2023) → 5.05% (2024) → 10.05% (2025).
- AI‑based automation (RPA): 3.38% (2023) → 4.41% (2024) → 6.68% (2025).
- Machine learning: 2.96% (2023) → 3.91% (2024) → 5.62% (2025).
Executive reading: the first wave centres on text and language, where data friction is low and return is visible. As firms consolidate data governance and connectors, automation and ML for analysis scale next.
For a professional services firm in Barcelona (around 60 employees), a realistic roadmap could be: 1) 2023–2024, NLG pilots for proposals and text mining on tickets; 2) 2024–2025, AI‑assisted back‑office automation and first ML models for prioritisation. The lesson is straightforward: progress in layers, measuring each milestone with time, quality and user‑satisfaction metrics.
To speed up capability building, structure enablement with targeted AI and Claude training for businesses focused on real data and role‑specific workflows.
Priority AI use cases and how to choose them well
Choosing the wrong first AI use case wastes months; choosing well unlocks momentum. Apply these criteria:
- Volume and repetition: frequent tasks with tacit rules or decision patterns.
- Accessible data: text/documents/tickets with identifiable histories.
- Low/medium risk: avoid starting with high‑stakes decisions.
- Clear metrics: time per task, error rate, SLA, unit cost.
Typical cross‑functional candidates:
- Commercial/marketing: proposal drafting, email personalisation, lead categorisation, feedback analysis.
- Customer service: initial answers with NLG/voice, conversation summarisation, smart routing.
- Operations/quality: data extraction from delivery notes/orders, duplicate/incident detection, prioritisation.
- HR/legal: CV screening, job‑description drafting, clause extraction.
How to turn them into executable projects:
- Define the target metric (what you will improve). 2) Specify the ideal flow (inputs, outputs, validations). 3) Bound the dataset for training or evaluation. 4) Pick the technology (NLG/NLU, AI‑based RPA, ML). 5) Design human‑in‑the‑loop controls and fallbacks. 6) Close the measurement with a verifiable before/after.
If you need help to select high‑value, low‑risk starters and to assemble a realistic roadmap, consider an AI consulting assessment and roadmap. This ensures governance, security and training are addressed from day one and that each pilot is framed to produce measurable business outcomes.
From diagnosis to first results: a 90‑day delivery plan
A 90‑day approach balances ambition with realism.
Weeks 0–2: diagnosis and roadmap
- Process and data mapping; interviews; risks and dependencies.
- Select 2–3 use cases with verifiable ROI and different risk levels.
- Governance plan (roles, security, privacy) and enablement plan.
Weeks 3–6: pilots and quick wins
- Configure prototypes with NLG/NLU and AI‑based RPA in real flows.
- Establish baseline metrics and human validation circuits.
- Document findings, data frictions and integration requirements.
Weeks 7–12: industrialisation and scale‑up
- Integrations with corporate tools; traceability and auditability.
- A/B prompts/flows; harden SLAs and roll‑out by team.
- Decision forum with time, quality and cost metrics.
To accelerate, review ready‑to‑deploy AI agents for business that integrate with CRM, ITSM or knowledge repositories. ARTIC often co‑implements with in‑house teams to set objectives, measure impact and transfer capabilities without creating vendor lock‑in.
AI agents and practical enterprise architecture
AI agents orchestrate tasks across tools (CRM, ERP, email, knowledge bases) and execute semi‑autonomous workflows with human control. In practice, a “read‑reason‑act” agent combines: 1) context understanding (documents, tickets, orders), 2) reasoning with business policies, 3) actions in systems (create/update records) and 4) reporting with traceability.
Design guidelines:
- Start with “assisted” agents (human approval) before moving to automatic actions.
- Break work into observable, testable steps; log evidence for each decision.
- Limit scope by data and permissions; record every action with audit trails.
- Measure precision/recall by task type, plus time and rework metrics.
Architecture notes: agents benefit from a clear knowledge base, event instrumentation, scoped credentials, and safe tool‑use policies. When your organisation requires enterprise‑grade governance and secure collaboration around advanced models, see how we operate as part of the Claude ecosystem via ARTIC’s Claude Partner profile. This helps align model choice, deployment and governance with your compliance posture and risk appetite.
How to measure AI ROI: metrics that matter
Without metrics, AI remains a perpetual demo; with metrics, it becomes a business lever. Define a scorecard per use case that separates direct impact, quality and adoption.
- Direct impact: average time per task, throughput, unit cost, backlog.
- Quality: accuracy, rate of critical errors, deviations vs. policy.
- Adoption: % of tasks touched by the assistant/agent, internal user satisfaction.
- Risk/compliance: audit coverage, share of actions with full traceability.
Methodology: establish a 4–6 week baseline; instrument the solution to capture events; then compare with an equal period after deployment. Segment by task type and complexity. Keep a “safety margin” for sensitive decisions (confidence threshold + human review).
Practical examples: in operations, measure time reduction in customer onboarding or reconciliations; in customer service, first‑response time; in commercial, the number of proposals sent per week without worsening win rate. Remember to track the cost of guardrails (human checks, rework) to get a full view of ROI.
People, training and governance: the differentiator
AI technology pays off in proportion to team adoption. Every deployment should be paired with a “people and governance” plan.
- Roles and responsibilities: product owner per use case; data steward for quality and access; a security/compliance lead; and an AI committee to prioritise and arbitrate risks.
- Enablement: train teams to use assistants effectively, design prompts and validate outputs; train managers on metrics and governance.
- Security and privacy: data‑handling policies, permitted‑use lists, per‑agent permissions and audit logs.
- Cultural change: recognise improvements, share learnings and iterate on real user frictions.
Practical training plan:
- Baseline enablement for all users (2–4 hours) focused on daily tasks and quality criteria. 2) Function‑specific workshops (sales, support, operations) on real data. 3) Manager training on metrics, risks and scale‑up. Align incentives so that time saved is reinvested in higher‑value work. If you need structured curricula and certification by role, a programme like AI and Claude training for businesses can anchor consistent practices across teams while keeping governance requirements front and centre.
Frequently asked questions
What share of companies in Spain use AI, and how does that compare with the EU?
In 2025, 20.27% of Spanish enterprises with 10 or more employees used at least one AI technology, versus 19.95% in the EU‑27. The figures come from Eurostat’s ICT usage in enterprises survey (dataset isoc_eb_ai, updated 2026‑06‑15).
Do large enterprises use AI more than SMEs?
Yes. In Spain, 57.46% of firms with 250+ employees use AI, compared with 30.1% of those with 50–249 and 17.17% of those with 10–49. The EU‑27 shows a similar pattern with 55.03%, 30.36% and 17% respectively (2025).
Which AI technologies are most common in companies?
In Spain in 2025, the most widespread are language/code generation (8.92%), text mining (10.05%), AI‑based automation (6.68%) and machine learning (5.62%). These are Eurostat figures for companies with 10+ employees, excluding the financial sector.
Has AI use grown in Spain in recent years?
Yes. The share of firms using at least one AI technology rose from 9.18% in 2023 to 11.31% in 2024 and 20.27% in 2025. The fastest gains are in text‑ and language‑based use cases (Eurostat, isoc_eb_ai).
How should a company start an AI project?
Pick a use case with accessible data, controlled risk and clear metrics. Run a 6–8 week pilot with human validation, measure before/after, and plan integration and scale‑up only if the target is met.
How do you measure the ROI of AI?
Create a use‑case scorecard with impact (time, cost), quality (accuracy, critical errors) and adoption (share of tasks). Establish a pre‑deployment baseline and compare it with the post‑deployment period.
What role do AI agents play in the enterprise?
AI agents orchestrate tasks across systems (CRM, ERP, email) and execute semi‑autonomous workflows under human control. They can accelerate responses, reduce repetitive work and add traceability when designed with clear scopes and metrics.
Sources and methodology
- Eurostat, survey on ICT usage and e-commerce in enterprises: companies with 10 or more persons employed, excluding the financial sector (database).
- The linked client examples are real ARTIC projects. Calculators are indicative.
