Aiolum Browse stories
All stories
AI document parsing and data extraction platform with visual citations · Spain

anyformat: Product, team and how it was built

Small teamFundedB2BDocument processing

anyformat (AI document parsing and data extraction platform with visual citations, Spain): 1 million documents processed, June 2026. Evidence grade E: no revenue figure was checked. Team: 2 co-founders; 9 people in November 2025, after a EUR 3.3 million seed round (The Officer). Launched June 2024; still running.Added 5 November 2026

Outcome
1 million documents processed, June 2026
Evidence
E no revenue figure was checked
Launch
June 2024
Status
Still running
Team
2 co-founders; 9 people in November 2025, after a EUR 3.3 million seed round (The Officer)
Outside funding
EUR 520,000 pre-seed, then a EUR 3.3 million seed round led by Kibo Ventures, November 2025 (as reported by The Officer)
Group
Documents, data and slides
Main channel
Not stated
No revenue figure

What is not verified

No revenue figure was found. We looked for a payment-processor profile, a founder interview with revenue, and a company-register filing, and found none. The Officer wrote in November 2025 that the company's goal is EUR 1 million of annual recurring revenue in 2026; that is a target, not a result. The headline, 1 million documents processed, rests on one sentence in an Up Next event report of 6 June 2026 that does not give its source. A reader could look for the company's annual accounts in the Spanish companies register or a later funding announcement with figures. This story is grade E: the business is real and its product, price and customers are described from public pages, but no revenue figure was checked. Aiolum shows it as an idea worth studying, not as proof that it makes money.

anyformat is an intelligent document processing (IDP) platform: software that reads business documents and writes the data in them into fields that other systems can use. Each value comes with a confidence score, a number that estimates how likely the value is correct, and a visual citation, a link to the exact place on the page where the value was found. The company's pricing page in November 2026 lists a free plan with 50,000 credits, a Business plan at EUR 499 a month, and custom Enterprise contracts.

Juan Huguet and Diego Pérez, who both worked at Clarity AI before, founded the company in Madrid, according to The Officer in November 2025. The earliest archived copy of the site, from June 2024, shows a "Gen-AI data extraction platform" with a demo-request form. The same article reported a EUR 520,000 pre-seed round, a EUR 3.3 million seed round led by Kibo Ventures, and a team of nine people.

No revenue figure was found. A South Summit Madrid 2026 report by Up Next, dated 6 June 2026, describes anyformat as having processed 1 million documents, with customers including L'Oréal, the Government of Singapore and Iberia; that is the headline on this page, and it was not checked against any other source.

anyformat: 1 million documents processed, June 2026. AI document parsing and data extraction platform with visual citations, Spain, team of 2 to 10, evidence grade E.
The story in one card. Share it with the address of this page.
01

How much revenue does anyformat make?

Each row shows the evidence grade and the source numbers from the list at the end of this page. The grades are explained in How Aiolum verifies.

A funded two-founder company in Madrid with a team of nine in November 2025, with no revenue figure published
FactValueGradeHow we knowSources
Documents processed, June 20261 millionDUp Next report on South Summit Madrid 2026, 6 June 2026: "1M docs processed". The report does not say who gave the figure or how it was checked.[1]
Named customersL'Oréal, IAG, Iberia, the Government of SingaporeCThe CEO's author page on the company site; Up Next names L'Oréal, the Government of Singapore and Iberia. No customer statement was found.[6, 1]
FoundersJuan Huguet (CEO) and Diego Pérez Sastre (CTO), both formerly at Clarity AICThe Officer: "respondieron Juan Huguet y Diego Pérez, ex Clarity.ai, cuando fundaron Anyformat" (Juan Huguet and Diego Pérez, formerly of Clarity.ai, answered that question when they founded Anyformat) (translation). The company's author pages give the roles.[2, 6, 7]
Team, November 20259 peopleCThe Officer, 27 November 2025: "hoy con un equipo de nueve personas" (today with a team of nine people) (translation). The careers page in November 2026 says "a small, high-caliber team" with hubs in Madrid and Palma de Mallorca.[2, 8]
FundingEUR 520,000 pre-seed, then EUR 3.3 million seedCThe Officer reported that the seed round was led by Kibo Ventures with 4Founders, Abac Nest Ventures and Decelera Ventures. The company's press page lists an Expansión article with the same EUR 3.3 million figure; Expansión blocks AI agents, so only its headline was read, on the press page.[2, 5]
Earliest archived siteJune 2024CInternet Archive copy of 18 June 2024: "The Gen-AI data extraction platform", with a demo-request form and no prices. An archived redirect from April 2024 exists.[9]
Prices, November 2026free plan with 50,000 credits; Business EUR 499 a month; Enterprise customCCompany's pricing page.[4]
Security certificationISO 27001:2022CCompany blog, April 2026: the audit was made by Prescient Security. The company says it runs on AWS with EU data residency.[11]
Models usedseveral providers; the company calls itself model-agnosticCCTO's blog post of 16 June 2026: its pipelines give the same output "whether they're running on Claude, GPT, Gemini, or something that doesn't exist yet".[10]
Time from the first archived site to the June 2026 figure24 monthscalcCalculated: June 2024 to June 2026.[9, 1]

How to read this table: grade C is the company's own page or a journalist's article; grade D is a figure from one third-party report that does not say how it was checked. No value is grade A or B, and no row is a revenue figure.

anyformat prices read on the company's pricing page on 5 November 2026, in euros, not converted to US dollars (grade C)
PlanPer month, billed yearly (US$)Per month, billed monthly (US$)Limit
Pay As You Go0050,000 credits once, then pay as you go; 1 user, 2 production workflows
Businessabout 399 (calculated)499500,000 credits a month; unlimited users; ISO 27001; private Slack channel support
Enterprisecustomcustomself-hosted deployment, custom contracts and SLAs, dedicated support

All values in this table are euros. The pricing page has a monthly/annual switch that says "Save 20%"; the annual price is loaded by script and was not visible in the downloaded page, so the yearly value for Business is calculated on this page as EUR 499 less 20%. Credits pay for the operators used. Extra credits cost EUR 0.0015 each on the free plan and EUR 0.001 each on Business; Enterprise gets a volume discount.

02

When did anyformat launch, and how did it grow?

From an archived demo page in June 2024 to a seed round and an ISO 27001 certificate. Each entry carries its grade.

  1. June 2024CThe earliest archived home page presents a Gen-AI data extraction platform for PDFs, emails, text, images and audio, with a demo-request form. [9]
  2. 2025CThe co-founders publish blog posts on the MCP protocol for AI agents (May 2025) and on scoring the confidence of structured AI output (June 2025). [7, 6]
  3. 27 November 2025CThe Officer reports a EUR 3.3 million seed round led by Kibo Ventures, after a EUR 520,000 pre-seed, and a team of nine people. [2]
  4. April 2026CISO 27001:2022 certificate, audited by Prescient Security. [11]
  5. 6 June 2026DUp Next lists anyformat among the standout startups at South Summit Madrid 2026, with 1 million documents processed. [1]
  6. November 2026CPublic prices with a free plan and a EUR 499 Business plan; home page benchmarks against GPT, Claude, Gemini and OCR services. [3, 4]
03

Why did anyformat work?

Four causes, in the order they mattered. Each one is a reusable lesson.

01

Sell the check, not the reading

General AI models already read documents. anyformat's pages put the weight on the parts a company needs to trust the output: a calibrated confidence score per field, a citation to the page position, rules written in plain language, and a person who reviews only the flagged fields.

02

Long documents in one pass

The company says its extraction works across hundreds of pages and long tables without splitting the file into pieces, and its home page shows its own benchmark on documents of up to 50 pages. These are the company's own tests.

03

Enterprise requirements early

EU data residency, self-hosted and air-gapped deployment, and an ISO 27001 certificate are on the home page. These are the conditions that banks, insurers and governments ask for, which is where its named customers are.

How it was built

These details come from the company's own pages and blog. They are grade C. [10, 11, 12, 3]

  • Five operators (parse, extract, classify, split, validate) connected in a typed workflow graph that runs in one API call, with an SDK, a command-line tool and a connection for AI agents.
  • Model-agnostic design: the CTO writes that the pipelines run on Claude, GPT or Gemini and keep the same confidence scores and audit trails.
  • Hosted on AWS with keys in AWS KMS, AES-256 encryption at rest and TLS 1.2 or higher in transit.
  • ISO 27001:2022 audit by Prescient Security in 2026.
  • The CTO wrote in January 2026 that he built slide decks with Lovable and Claude Code instead of slide software.

Tools named in the sources

Each one opens its page in the Tools tab, with every story that uses it.

GPTCGeminiCClaude (assistant)CLovableCClaude CodeCNext.jsBAmazon Web Services (AWS)CGoogle Tag ManagerBPrescient SecurityCSlackC
04

Who tried this before anyformat, and why did they fail?

Four companies in our database of failed startups promised to take the work out of paperwork or to turn messy sources into structured data. The values below are from that database and are not independently verified.

Died 2011 to 2017Burned US$17M

Canvas

The promise

Turn paper forms into digital workflows.

Why it died

Microsoft and Adobe offered the same inside their suites, and low-code platforms took the rest.

Competition
Died 2016 to 2019Burned US$500K

Lendsnap

The promise

Gather a borrower's mortgage documents automatically.

Why it died

One document feature was not enough against Blend's all-in-one mortgage platform.

Competition
Died 2019 to 2022Burned US$3M

Dashblock

The promise

Turn any web page into an API with no code.

Why it died

Its technology fell behind AI methods and the constant changes of web pages.

Competition
Died 2012 to 2023Burned US$852M

Olive AI

The promise

Automate the administrative work of hospitals, such as prior authorizations and billing.

Why it died

The automation needed much human work behind it, and costs grew faster than revenue.

Unit Economics

The failed startups come from our database of 1,749 dead startups, retold here in our own words. Its entries are AI-assisted summaries of public sources, so these cards carry no evidence grade.

What changed

Canvas and Lendsnap sold one step of paperwork and lost to larger suites that added the step for free. Dashblock and Olive AI promised automation that, in the database's account, still needed manual repair. anyformat faces the first risk directly: the model makers and the cloud providers (AWS Textract, Google Document AI, Azure Document Intelligence) all sell document reading.

The difference anyformat claims is that large language models can now read almost any layout without templates, so the hard part has moved from reading to proving the result is right. Its product is built around that proof (confidence, citations, validation rules, audit trails), which the general models do not provide by themselves.

05

Can one person copy anyformat now?

Not as a solo project. The idea of selling verification, yes.

anyformat is two founders with physics and AI-engineering backgrounds, a team of nine and EUR 3.8 million of reported funding, selling to large enterprises with security certificates. One person cannot copy that sales motion. The narrower idea is open: a small tool that adds confidence scores and page citations to AI extraction for one document type and one industry.

Where it is still open: Single-industry extraction with audit trails: customs documents, insurance claims, medical invoices for billing teams, public tenders.

  1. 1Pick one document type that a regulated team checks by hand, such as bank statements or bills of lading.
  2. 2Return every value with its page position and a confidence score, and send only low-confidence fields to a person.
  3. 3Publish a benchmark on real documents against the general models, and repeat it when new models appear.
  4. 4Offer a free credit allowance so developers can test before sales calls.

What could go wrong for anyformat?

The risks are the mirror image of the causes above.

What is not verified

No revenue figure was found. We looked for a payment-processor profile, a founder interview with revenue, and a company-register filing, and found none. The Officer wrote in November 2025 that the company's goal is EUR 1 million of annual recurring revenue in 2026; that is a target, not a result. The headline, 1 million documents processed, rests on one sentence in an Up Next event report of 6 June 2026 that does not give its source. A reader could look for the company's annual accounts in the Spanish companies register or a later funding announcement with figures.

Customers and testimonials

The CEO's author page names L'Oréal, IAG, Iberia and the Government of Singapore as customers, and Up Next repeats three of them; no customer statement was found. The home page quotes a testimonial attributed only to "AP Operations Lead, Global manufacturer". The Spanish press articles of May 2026 listed on the press page discuss document automation in general and give no figures about the company.

Competition

The company's own pages compare it with more than 30 products, from AWS Textract and Google Document AI to Reducto and LlamaParse. On this site, PDF Vector (a document parsing API, grade C) and Normain (cited answers from documents, grade E) sell similar products to smaller customers.

Dependence on model providers

anyformat's home page benchmarks itself against GPT, Claude and Gemini, which it also uses. Its CTO wrote on 16 June 2026, after a government order took two Anthropic models offline, that the pipelines run on Claude, GPT or Gemini and are built to survive a model swap. If the general models add calibrated confidence and citations themselves, the gap anyformat sells becomes smaller.

Team size

The company was started by two co-founders and had nine people in November 2025 after a seed round, according to The Officer. The careers page lists open roles and hubs in Madrid and Palma de Mallorca, so the team may be larger now.

More in Documents, data and slides

Other businesses in the same group, best evidence first. All 46 stories in the group, 14 with grade-A revenue.

AI presentation generator and Google Slides add-on · India

MagicSlides

Evidence A
$723K total, now $9K/mo

Team of 2 to 10 · 45 months after launch, revenue falling

Said US$19K a monthRecords US$9.6K in JulyDid not hold

It appeared in January 2023 inside Google Slides, a program where people already work, when writing slides from one sentence was new, and users showed it to each other without paid advertising. A blog about presentation topics then brought about 80% of new users from Google search, by the founder's account. Both channels weakened: search traffic fell, and Google added the same function to Slides. The verified total is real, and so is the fall: monthly revenue fell by half between February and July 2026.

7 grade-A facts · 4 that failed before it
AI presentation generator (web app) · USA

GenPPT

Evidence A
$110.6K in 12 months

Solo founder · 29 months after launch, revenue falling

It is run by a search marketer: its pages target what people type into Google, and its best month came from a list article on another company's blog that ranked there. A US$1 trial that needs a card turns visitors directly into subscribers at US$25 a month or US$99 a year.

7 grade-A facts · 3 that failed before it
Step-by-step guides and interactive walkthroughs from screen captures, with AI writing, voiceover and translation · India

Guidejar

Evidence A
$85.1K in 12 months

Solo founder · 36 months after launch, revenue flat

The first customers came from a lifetime-deal sale on AppSumo, at a time when the product had no AI at all. That gave about a thousand users, word of mouth and a Powered by Guidejar link on every free guide. From April 2024 the founder sold subscriptions instead, and added AI writing, voiceover and translation to the paid plans, metered in credits.

5 grade-A facts · 4 that failed before it
AI presentation generator · United Kingdom

Presentia AI

Evidence A
$60K total, now $1.1K MRR

Team of 2 to 10 · 24 months after launch, revenue falling

It sold a simple promise, a finished PowerPoint file from one sentence, at a low price: in 2024 and the first half of 2025 the archived pricing showed unlimited decks for US$9.99 a month. Stripe payments read by TrustMRR add up to US$60,120 since the start. The verified monthly figures also show a steady fall after November 2025, when the plans had moved from unlimited use to slide credits, and the founder put the business up for sale.

9 grade-A facts · 4 that failed before it
AI functions for Google Sheets (add-on) · India

SheetAI

Evidence A
$52K total, now $367 MRR

Team of 2 to 10 · 47 months after launch, revenue falling

It put a language model inside Google Sheets from November 2022, and Google's marketplace brought more than 201,000 installs. Few installs paid: US$52,136 of verified revenue in 47 months is about US$0.26 per install. Its most important result may be indirect: a request for a pitch deck for SheetAI led the founder to build MagicSlides, which by its own verified data earned about 14 times as much. SheetAI's own revenue is now falling, and Google has offered a similar AI function inside Sheets since June 2025.

5 grade-A facts · 4 that failed before it
AI agent inside Excel · France

ElyxAI

Evidence A
$2.8K MRR, listed for sale

Solo founder · 15 months after launch, revenue falling

It does the work in the cells instead of explaining what to paste, and it works in the Excel that people already have, without a Microsoft Copilot license. The founder already taught Excel to a large French audience. The AI models are bought by the request from OpenAI and Anthropic, so one person could build and run it.

9 grade-A facts · 5 that failed before it

Built with the same tools

Businesses whose sources name tools that anyformat also uses: Gemini, Next.js, Google Tag Manager, Slack. Every tool has its own page in Tools.

AI search visibility analytics for marketing teams and agencies · Germany · Funded

Peec AI

Evidence B
$10M ARR in 15 months

Started small, grew after funding · 15 months after launch, revenue growing

Three founders who met in an Antler cohort in Berlin launched it in February 2025, in a category that, as The Next Web wrote in May 2026, "barely existed 18 months ago". With self-serve plans and unlimited users it signed 1,300 brands and agencies in about nine months, the company said, and two funding rounds within ten months paid for a large sales and engineering team in Berlin and New York.

Every number graded · 3 that failed before it
Desktop workspace for AI coding agents · United States

BridgeMind

Evidence C
$251K/yr run rate

Solo founder · 14 months after launch, revenue growing

The founder builds the product live on YouTube, X and Twitch, four to six hours at a time on several days a week, and the title of each stream shows the current annual run rate. The viewers are the target customers: people who write software by giving instructions to AI agents. A reviewer who watched one stream says that every paying customer came from streaming and that nothing was spent on ads.

Every number graded · 5 that failed before it
AI platform that chases late B2B invoices · France · Funded

Cleavr

Reported, not verified
$1.12M ARR reported

Team of 2 to 10 · 7 months after launch

Chasing late invoices is repetitive work that finance teams do by hand in spreadsheets and email. Cleavr connects to tools that companies already use, such as Pennylane, Stripe and SAP, so a customer can start without a migration, and the company says setup takes under a day.

1 grade-A fact · 4 that failed before it
Local SEO software for agencies: rank tracker, directory listings and AI-written articles · Canada

LocalRank

Evidence A
$730K total, now $21K/mo

Team of 2 to 10 · 19 months after launch, revenue falling

Said 500+ paying agenciesRecords 207 subscriptionsDid not hold

The founder had an audience before the product existed: a daily YouTube show about SEO since 2023 and an X account that has 62,000 followers today, so the first customers cost nothing. A developer co-founder built the software with AI coding tools. The product started as one plain rank tracker at US$19 a month, grew into a bundle at US$57 to US$2,997 a month, and by July 2025 the first payment of a new customer from a paid ad covered the cost of that ad.

5 grade-A facts · 5 that failed before it

Sources

The grade next to each source is the best grade that the source can support. "Read in full" means that the text of the page was downloaded and the numbers were checked in it.

  1. D Up Next, 6 June 2026: South Summit Madrid 2026: Mitiga Solutions Wins, Plus the Standouts
    Read the section on anyformat. Source for the 1 million documents and the three named customers.
  2. C The Officer, 27 November 2025: Anyformat, la startup que ordena el caos documental con inteligencia artificial
    Read in full (Spanish). Source for the founders, the pre-seed and seed rounds, the team of nine and the 2026 revenue target.
  3. C anyformat home page, read 5 November 2026
    Read in full. Source for the product, the benchmarks, the security claims and the testimonial.
  4. C anyformat pricing page, read 5 November 2026
    Read in full. The annual price is loaded by script and was not visible.
  5. C anyformat press page, read 5 November 2026
    Read in full. Lists the Expansión funding article (not opened: Expansión's robots.txt disallows ClaudeBot) and the May 2026 Spanish articles.
  6. C anyformat author page: Juan Huguet García, CEO and co-founder
    Read in full. Source for the CEO's background, the investors and the named customers.
  7. C anyformat author page: Diego Pérez Sastre, CTO and co-founder
    Read in full. Source for the CTO's background and the list of blog posts with dates.
  8. C anyformat careers page, read 5 November 2026
    Read in full. Source for the small team and the office hubs.
  9. C Internet Archive copy of anyformat.ai, 18 June 2024
    Read in full. The earliest archived home page.
  10. C anyformat blog, 16 June 2026: Your AI Stack Can Disappear Overnight. Now What?
    Read in full. Source for the model-agnostic design and the models named.
  11. C anyformat blog, April 2026: ISO 27001:2022, Certified. The Trust Was Engineered Before the Audit.
    Read in full. Source for the auditor, AWS KMS and the encryption details.
  12. C anyformat blog, 22 January 2026: AI Agents Don't Kill Document Processing. They Make It Inevitable.
    Read in full. Source for the use of Lovable and Claude Code.
  13. C Database of failed startups, local export
    Local export. AI-written summaries, not independently verified. Source for the failed predecessors.

Run this company? Send corrections or a verified figure to aiolum@livyden.com

Back to all stories