The unglamorous truth about enterprise AI is that most of the value still sits in PDFs nobody wants to read. Invoices, contracts, customs paperwork, the kind of documents that pile up in a shared drive and quietly cost a finance team its evenings. Anyformat, a Madrid outfit founded last year, has spent eighteen months pitching large institutions on the idea that an agent can read those documents, cross-check them against an internal database, and route the result, without sending anything to a US hyperscaler.
It is a narrow bet, sensibly scoped, and the early customer list is more interesting than the round size suggests.
A €3.3M seed with a sovereignty pitch
On 27 November, Anyformat closed a €3.3 million seed led by Kibo Ventures, with 4Founders, Abac Nest Ventures, and Decelera Ventures returning from the company's €520,000 pre-seed in June [Yahoo! Finance, November 2025] [EU-Startups, November 2025]. The capital is earmarked for engineering hires and a proprietary error-detection model, plus a commercial push across Europe [PRNewswire, November 2025].
Pre-seed (Jun 2025) | 0.52 | M USD
Seed (Nov 2025) | 3.57 | M USD
No valuation was disclosed, which is normal at this stage and not worth dwelling on. The number that matters more is the one in the forward guidance: management expects to cross €1 million in ARR in 2026 [Yahoo! Finance, November 2025]. For a company that closed its first institutional cheque five months ago, that is a brisk pace, assuming it lands.
What the agent actually does
The product reads contracts, invoices, emails, and presentations in any layout, extracts structured data, and triggers downstream workflows like approvals or routing without a human in the loop [anyformat.ai]. It combines generative models with deterministic rules and a no-code orchestration layer, which is how the company avoids the brittleness that has dogged pure-LLM extraction pipelines [PERPLEXITY SONAR PRO BRIEF].
The wedge the founders point at is real-time fusion with a customer's own database. Anyformat says it can cross-reference an extracted invoice line against internal catalogs, policies, or historical records as the document is being processed [PERPLEXITY SONAR PRO BRIEF]. That is the difference between a tool that gives you a JSON blob and one that can tell a finance team whether the invoice should be paid at all.
The customers doing the validating
The reference list is the most persuasive part of the story. Anyformat has deployments at L'Oréal Group, IAG Group, Iberia, and the Singapore Government [Cardboard, 2026] [Pulse 2.0, November 2025]. On L'Oréal's commercial invoicing team, the company reports a 60% reduction in processing time and 99% extraction accuracy [anyformat.ai]. Those are company-supplied figures, but the named buyer is hard to fake.
The Singapore engagement came through Dimension X Cohort 5, run by HTX, the Home Team's science and technology arm. Anyformat worked on proof-of-concept tooling for public safety and security applications [SGInnovate, July 2025]. That is a long way from L'Oréal's accounts payable, which tells you something about how horizontal the underlying engine is, and how much sales-cycle risk the company is carrying in parallel.
The competitive read
Document intelligence is one of the most crowded categories in enterprise software. ABBYY has been at this since the 1990s. Hyperscience and Rossum each raised growth rounds well north of $100 million. Amazon Textract is bundled into AWS and prices like a commodity. IBM Datacap and Laserfiche own large parts of the regulated-enterprise installed base. Anyformat is entering with a seed round that would not cover a quarter of Hyperscience's sales payroll.
The four reasons the bet is still defensible:
- European data sovereignty. EU public-sector and regulated buyers are increasingly unwilling to route documents through US-hosted IDP stacks, and Anyformat's pitch is built around that constraint [EU-Startups, November 2025].
- Agentic workflows, not just extraction. The product triggers approvals and routing natively, where most legacy IDP tools stop at the JSON handoff [anyformat.ai].
- Real-time database fusion. Cross-referencing extracted fields against a customer's proprietary records during processing is a genuine architectural difference [PERPLEXITY SONAR PRO BRIEF].
- Named early references in two continents. L'Oréal, IAG, Iberia, and HTX provide validation in finance, aviation, and public safety, three verticals with very different document grammars [Pulse 2.0, November 2025].
The honest counter is that ABBYY and Rossum can both credibly claim European data residency, and Amazon Textract competes on price in a way a Madrid seed-stage company cannot. The €3.3 million has to buy enough engineering velocity to keep the agentic and database-fusion gap open for at least two product cycles.
The team
| Name | Role | Source |
|---|---|---|
| Juan Huguet-García | Co-founder, CEO | [LinkedIn, 2026] |
| Diego Pérez-Sastre | Co-founder, CTO | [EU-Startups, November 2025] |
Two founders, no disclosed prior exits in the cited coverage, and a backer roster that has now written two cheques in five months. Kibo Ventures leading the seed after sitting in the pre-seed is the strongest internal signal in the round.
What to watch in 2026
The €1 million ARR target is the headline number, but the more telling milestone will be whether Anyformat converts the HTX proof-of-concept into a paying public-sector contract, and whether it lands a second flagship in regulated European finance or insurance. A Series A in the next twelve to eighteen months is the obvious follow-on, and the size of it will reveal what enterprise renewal rates actually look like.
A back of envelope on the ARR bet: €1 million across, say, ten enterprise accounts implies roughly €100,000 per customer per year. That is in the same neighborhood as a mid-market Rossum contract and well below what Hyperscience charges its Fortune 500 base. If Anyformat can hold that price point while pointing at L'Oréal in every deck, the unit economics start to look interesting. If it has to discount to displace Textract on the long tail, less so.
The company Anyformat most needs to beat is Rossum. Both are European, both pitch agentic document workflows, both court the same data-sovereignty-anxious finance buyer. Rossum has the head start and the larger balance sheet; Anyformat has a tighter product story and the willingness to walk into a Singapore Home Team lab. The next round will tell us which of those mattered more.
Sources
- [Yahoo! Finance, November 2025] Anyformat closes a €3.3 million seed round led by Kibo Ventures | https://finance.yahoo.com/news/anyformat-closes-3-3-million-110000652.html
- [EU-Startups, November 2025] Madrid's Anyformat closes €3.3 million round | https://www.eu-startups.com/2025/11/madrids-anyformat-closes-e3-3-million-round-to-scale-its-agentic-ocr-and-data-sovereignty-focused-document-tools/
- [Pulse 2.0, November 2025] Anyformat Raises €3.3 Million Seed Funding | https://pulse2.com/anyformat-3-3-million/
- [PRNewswire, November 2025] Anyformat closes a €3.3 million seed round | https://www.prnewswire.com/news-releases/anyformat-closes-a-3-3-million-seed-round-led-by-kibo-ventures-with-the-aim-of-transforming-document-management-for-global-corporations-302627516.html
- [anyformat.ai] AI Document Processing & Workflow Automation | https://anyformat.ai/
- [Cardboard, 2026] How anyformat automated 95% of SaaS expense management | https://cardboard.inc/customers/anyformat-ai/
- [SGInnovate, July 2025] Dimension X Cohort 5 Demo Day | https://www.sginnovate.com/event/dimension-x-cohort-5-demo-day-hatch-innovation-centre-htx
- [LinkedIn, 2026] Diego Perez Sastre | https://www.linkedin.com/in/diego-perez-sastre/
- [PERPLEXITY SONAR PRO BRIEF] Web-grounded research compilation