Artificial intelligence has moved beyond the status of a technological promise to become a concrete and measurable lever for value creation. The question for executives is no longer whether to integrate AI, but how to do so in a way that generates sustainable competitive advantage while minimising regulatory and operational risk. In the Spanish business context of 2024–2025, this challenge unfolds alongside the entry into force of the European AI Act — the world's first comprehensive AI regulatory framework.
The Real State of AI Adoption in Spanish Companies
According to INE data for 2024, 38% of large Spanish companies (more than 250 employees) use at least one AI technology: big data analytics, machine learning, natural language processing or computer vision. However, the dispersion is enormous: financial services, insurance and technology companies show adoption rates above 60%, while construction, retail and hospitality are below 20%.
The 2024 report from Spain’s Centre for the Development of Industrial Technology (CDTI) highlights that the primary barrier is not the availability of the technology — large language models (LLMs) such as GPT-4, Gemini or Claude are accessible to any company by subscription — but the shortage of internal talent to identify use cases with a positive ROI, prepare the necessary data and integrate solutions into existing processes.
Use Cases with the Highest Return in Finance and Operations
Automation of Financial Processes
The combination of RPA (robotic process automation) with conversational AI has enabled mid-sized companies to cut the time spent on the monthly accounting close by between 30% and 50%. In accounts payable, intelligent data extraction models for invoices (smart OCR) eliminate manual data entry with accuracy rates above 95%, reducing processing costs per invoice from €8–12 to under €1 in high-volume installations. Business management software company Sage estimated in 2024 that Spanish SMEs that automate invoicing and bank reconciliation processes save an average of 12 hours of administrative work per week.
Predictive Analytics and Cash Flow Forecasting
Machine-learning-based cash flow forecasting models consistently outperform traditional regression models when fed sufficient historical data (a minimum of 24 months of transactions). In scenarios with high seasonality or variability, the reduction in forecast error at 90 days can exceed 40% compared with simple moving-average models. This has direct implications for working capital optimisation and for reducing the cost of short-term credit lines.
Due Diligence and Contract Analysis
The review of contract documentation using AI (contract analytics) is one of the highest-impact use cases in M&A and corporate finance transactions. Tools such as Luminance or Kira Systems can review hundreds of contracts in hours, identifying change-of-control clauses, confidentiality obligations, assignment restrictions and penalties. In complex due diligence processes, this reduces the time billed by senior lawyers by between 20% and 40%, with a direct impact on the total cost of the transaction.
Supply Chain Optimisation
Combinatorial optimisation and reinforcement learning models applied to demand planning and inventory management typically generate savings of 10–20% in capital tied up in stock for distributors handling more than 5,000 product lines. In the food distribution sector, several Spanish chains have reported reductions in fresh-food waste in excess of 25% through point-of-sale demand forecasting models.
Generative AI and Its Business Applications
Large language models (LLMs) represent the most recent and disruptive wave. Immediate business applications include:
Content generation and communications: drafting reports, executive summaries, commercial communications and value propositions in multiple languages. A consultancy firm can cut proposal preparation time by 60–70% while maintaining quality, redirecting the time saved to personalisation and differentiating strategic analysis.
Internal virtual assistants on corporate knowledge bases: retrieval-augmented generation (RAG) systems enable employees to query the compliance policy, procedures manual or historical case files in natural language without specific training. Deloitte reported in 2024 that this type of solution reduces new employee onboarding time by 25% and queries to HR and Legal by 30%.
Code generation and review: for companies with development teams, AI-powered coding assistants (GitHub Copilot, Amazon CodeWhisperer) increase developer productivity by between 30% and 55% on tasks involving writing new code and refactoring, according to field studies from 2023–2024.
The AI Act: Obligations for Spanish Companies
Regulation (EU) 2024/1689, known as the AI Act, entered into force on 1 August 2024 and is directly applicable in Spain without requiring transposition. The obligations calendar is as follows:
- August 2025: absolute prohibition on AI systems presenting unacceptable risk (subliminal manipulation, mass social scoring systems, real-time facial recognition in public spaces, with limited exceptions).
- August 2026: full application of obligations for high-risk AI systems, which include personnel selection systems, credit scoring, critical infrastructure management and systems used in the justice and security fields.
- August 2027: extension of obligations to general-purpose AI (GPAI) systems with systemic capabilities.
For most mid-sized companies, the focus should be on taking stock of all AI systems in use — including third-party SaaS tools — and classifying them under the AI Act’s risk taxonomy. High-risk AI systems will require technical documentation, conformity assessment, human oversight and registration in the EU database.
How to Build an AI Roadmap with Verifiable Returns
The most common mistake in corporate AI initiatives is to start with the technology rather than the business problem. An effective methodology follows these steps:
- Process mapping and pain quantification: identify the processes with the highest time consumption, highest error rate or greatest impact on customer satisfaction. Quantify the current cost of the problem.
- Bounded proof of concept (PoC): select the use case with the best impact-to-complexity ratio and develop a prototype over 4–8 weeks using real data in a controlled environment.
- ROI and scalability assessment: before committing to production investment, validate that the model performs with data outside the training sample and estimate the total cost of ownership (TCO), including model maintenance and updating.
- Deployment and change management: real-world AI adoption depends 70% on organisational change management, not on the technology itself. Involving teams from the PoC stage and communicating clearly the impact on roles is decisive for success.
At BMC we integrate technology into business strategy. Learn about our services at /en/corporativo/transformacion-digital.
