AI that actually works: the Data-ka case and automatic tender proposals

A Basque company uses NLP to read 200-page tender documents and generate proposal drafts. No hype, no smoke, just results.

---

I’m tired of AI articles that talk about “digital transformation,” “disruption,” and “the future of work” without giving a single concrete example. Today we’re going to talk about something real: a Spanish company using AI to solve a specific, measurable problem.

Data-ka is a Basque company specializing in Natural Language Processing. Their flagship use case: a system that reads public procurement documents, extracts key requirements, and generates proposal drafts that a human team then validates.

Nothing about “revolutionizing industries.” Just automating tedious work that consumes hundreds of hours.

The real problem

If you’ve worked with public administration or large companies, you know tender documents. Documents of 100, 200, sometimes 500 pages full of technical, administrative, and legal requirements. Reading them is days of work. Understanding them, weeks.

Companies that bid for contracts have teams dedicated to this. People whose job is to read tenders, extract requirements, and prepare proposals that meet all the points. It’s qualified but repetitive work. And it’s a bottleneck: you can’t bid on more contracts than your team can process.

Data-ka’s solution

Data-ka’s system does the following:

1. Document ingestion. Receives the tender in any format (PDF, Word, whatever) and processes it.

2. Requirements extraction. Using NLP, identifies mandatory requirements, evaluation criteria, deadlines, necessary certifications, etc.

3. Structuring. Organizes requirements into categories: technical, administrative, economic, solvency…

4. Draft generation. Produces a first proposal draft that covers all identified points.

5. Human validation. A team reviews the draft, adds company-specific information, and refines the document.

The result: what used to take a week now takes a day. And with fewer errors, because the system doesn’t skip requirements due to fatigue.

Why this is different from the hype

I like this case because it perfectly illustrates how AI should be used in business:

Concrete problem. Not “improving efficiency” in the abstract. Reading tenders faster. Period.

Clear metric. Processing time, requirements identified, proposals generated. You can measure whether it works or not.

Human in the loop. The system generates drafts, humans validate. It’s not “AI does everything,” it’s “AI does the tedious work so humans do the valuable work.”

Demonstrable ROI. If you process tenders in a day instead of a week, you can bid on more contracts. More bids = more business opportunities. The math works.

If you’re evaluating implementing AI in your company, read what nobody tells you about implementing AI in small business. Data-ka is an example of how to do it right.

What we can learn

1. NLP is mature for structured documents

Tenders are long documents but with predictable structure. There are sections for technical requirements, award criteria, administrative conditions… Modern NLP is very good at extracting information from these types of documents.

Other similar cases: contracts, financial reports, technical documentation, regulations. If you have long, structured documents that someone has to read repeatedly, there’s probably an automation opportunity.

2. The value is in the process, not the model

Data-ka hasn’t invented a new language model. They use existing technology (probably combinations of open source models and commercial APIs) applied to a specific business process.

The value isn’t in having “better AI” than the competition. It’s in understanding the problem well enough to design a process that works.

3. Human validation is key

The system doesn’t aim to eliminate humans. It aims to multiply their capacity. An expert who used to process 2 tenders per week can now supervise 10.

This is important because clients for these services (public administrations, large companies) need guarantees. They won’t trust a document generated 100% by AI. But they do trust a document generated by AI and reviewed by experts.

4. Spain has applied AI talent

Data-ka is Basque. They participate in BIND 2025 (SPRI’s industrial startup acceleration program). They have plans to expand to Europe and the United States.

Not everything is Silicon Valley startups. There are Spanish companies doing interesting things in AI, especially in industrial and process applications. This connects with what I wrote about AI adoption in Spain: the potential is there, it just needs execution.

Other real cases in Spain

Data-ka isn’t an isolated case. Some more examples:

Sherpa.ai (Basque Country): Federated AI that allows training models without moving data. Applications in banking and healthcare.

Inbenta (Barcelona): Semantic chatbots and search engines for customer service. They work with IBEX 35 companies.

Biometric Vox (Albacete): Voice biometrics for authentication. Used by banks and insurers.

Clarity AI (Madrid): AI for sustainability analysis and ESG investing. Valued at over 400 million.

These aren’t cases of “we’re going to be the next OpenAI.” They’re cases of “we’re going to solve a specific problem better than anyone using AI.”

The opportunity for Spanish companies

If you run a company or department, the question isn’t “should I use AI?” The question is “what process do I have that’s tedious, repetitive, and consumes qualified people’s time?”

The usual candidates:

• Reading and extracting information from documents
• Classifying emails, tickets, or requests
• Generating routine reports
• Answering frequently asked questions
• Visual quality control
• Contract analysis

You don’t need to build anything from scratch. There are companies like Data-ka that specialize in this. Or you can start with standard tools (Claude, GPT-5) and a well-designed process.

The key is starting with a concrete problem, clear metrics, and realistic expectations. Not with “digital transformation” and pretty PowerPoints.

---

Do you know other applied AI cases in Spanish companies? Have you automated any process with NLP? Share your experience.