A crooked invoice is a small insult to order. The vendor name is in one corner, the tax total in another, and a smudge has settled over the one number the accounting system actually needs. Humans can usually make sense of it. Software, which prefers its facts standing in neat rows, has historically been less amused. Alberto Gimeno built his career in the gap between those two worlds - and then discovered that parsing the document was only half the work. Someone also had to persuade companies to trust the result.
Gimeno came to that problem as a computer scientist and mathematician. At Universidad Autónoma de Madrid, he studied both subjects from 2009 to 2014, worked different jobs alongside his degrees, and twice represented the university in the Southwestern Europe Regional Contest for programming. The biographical shorthand is satisfyingly symmetrical: mathematics trained the appetite for structure, computer science supplied the means to impose it, and a pile of untidy business documents refused to cooperate.
The more important university souvenir was not a credential. It was a pair of collaborators. Gimeno met Carmelo Juanes Rodríguez and Adrián Ramírez del Río while all three were studying mathematics and computer science in Spain. They started building projects together, entered their first accelerator in 2014, and kept going through several companies and several changes of direction. In founder years, a decade is closer to a geological era. Their partnership survived it.
Three builders, one missing department
When the trio went all-in on Invofox, each founder was a backend engineer. Ramírez del Río became the machine-learning specialist. Juanes Rodríguez took charge of turning the technology into a usable product. That left a less comfortable question sitting on the table: who would sell it? Gimeno had some sales experience from services, but product sales demanded a repeatable motion rather than a friendly network and a bespoke proposal.
He took the commercial side. The transition sounds tidy only in retrospect. A technical founder has to exchange the pleasure of a solved problem for the ambiguity of another discovery call. Gimeno learned the sequence in detail: how to write the first email, qualify a prospect, run the call, demonstrate the product, present a proposal and follow up until the answer became clear. Later, that sequence could become a standard for new salespeople. Before it was a playbook, it was one founder doing every step.
“We meet on the intersection of both those areas, which is product.”Alberto Gimeno, on dividing leadership between go-to-market and engineering
The division was not a divorce from technology. Product became the common table. His co-founders owned the technical build; Gimeno owned distribution and go-to-market; each side brought evidence to decisions about what ought to exist. More recently, he has returned to hands-on coding with AI development tools. The boundary moved because the tools did. Good role design is less like carving names in stone and more like drawing a useful map in pencil.
How the founding team divides the work
Listen for the pull
The founders had previous experience with the opposite sensation. Gimeno described an earlier business as pushing a boulder uphill: progress was possible, but it demanded immense effort. Invofox began differently. Rather than construct a solution and hunt for an appreciative buyer, the team spoke with companies already struggling to turn PDFs and images into usable data. The same difficulty kept returning. Gimeno’s compact version is disarmingly modest: “We stumbled on a problem.”
They did not immediately pretend to know the ideal customer. The team tested three hypotheses at once: accounting and bookkeeping firms, construction and logistics companies with document-heavy operations, and software companies that needed document processing inside their own products. All three groups paid. All three wanted different things. Serving them together was painful, but the pain generated information.
Three hypotheses entered. One earned the focus.
Invofox remained spread across the three segments until it reached roughly $1 million in annual revenue. Then the founders committed their resources to software companies. Growth accelerated. The sequence matters. Focus did not arrive as a slogan pinned above a desk. It came after a deliberately messy experiment, at the moment when more exploration would have become an excuse not to choose.
For Gimeno, product-market fit is therefore tactile. Push feels like explaining the need over and over. Pull feels like meeting buyers who already know the pain and want to discuss the practical limits of a solution. Invofox’s prospects increasingly arrive after trying to build document extraction themselves. Their first attempt proves the concept. Production teaches them about the thicket.
The thousandth document is the test
A document demo can look miraculous. Upload a PDF, wait a beat, receive a tidy schema. Real documents, however, have a taste for variety: long mortgage files, handwritten notes, shifting formats, inconvenient tables, poor scans, mixed languages and fields that matter only under a customer’s private rules. The interesting engineering begins where the demonstration ends.
Gimeno argues that the distance between “works in a demo” and “works in production” is made mostly of failure modes rather than missing features. That belief explains why his sales conversations have become more technical and consultative. Prospects rarely need to be convinced that automated extraction is possible. They need to know what happens when the scan is faint, the format changes, the model updates, or a plausible number is wrong.
The answer lives around the model: validation architecture, business rules, exception workflows, integrations, monitoring and careful handling of data. Invofox’s pitch resembles the familiar case for payment infrastructure. A software company could build its own document-processing system, just as it could attempt its own payments stack. The real question is whether maintaining the long tail of complications is the best use of its engineers.
“People no longer doubt that it is possible.”Gimeno, on how the document-AI sales conversation changed
The company also points the technology inward. One internal AI system took over a monthly transactional billing process that previously occupied two people for 10 to 12 hours. Another helps the team find topics and develop content in individual voices. The detail is more persuasive than a sweeping promise because it is boring in exactly the right way. Useful automation often arrives disguised as the disappearance of an unpleasant spreadsheet ritual.
An ocean, with pets
In 2024, Gimeno moved from Madrid to San Francisco to lead the company’s US expansion. Corporate geography tends to be rendered as arrows on a map. His own account included the domestic cargo: his wife moved, their pets moved, and their daughter was born that year. Invofox opened a US office while keeping Madrid running. Juanes Rodríguez prepared to join him in San Francisco. The founding partnership was learning a new trick - operating across places without becoming two companies.
The move followed the same logic as the company’s market choice. Gimeno wanted proximity to software businesses that treat reliability, precision and trust as infrastructure rather than ornament. He has written that customers often begin with one document type and then find others: invoices lead to payslips, bank statements, contracts and mixed PDFs. His aspiration is explicit - he believes that expanding use inside customers can carry Invofox toward $100 million in yearly revenue. It is a target, not a present fact.
San Francisco did not erase Spain from the story. Gimeno continues to speak warmly about the Spanish accelerators, mentors and investors who helped the company. He has also volunteered with Tetuan Valley since 2015, lending a hand to its Startup School program. The Atlantic crossing enlarged the network; it did not require pretending the earlier one had been provincial.
A founder who kept changing jobs
There is a pleasant contradiction in Gimeno’s public interests. Y Combinator lists literature, CrossFit and baseball. One is built from sentences, one from repetitions, one from long periods of anticipation interrupted by precise action. Together they make a decent accidental syllabus for enterprise sales.
His career has the same mixed rhythm. He trained to build systems, then learned to sell them. He tested broadly, then chose narrowly. He left Madrid to become more present in San Francisco while keeping close ties to the people and institutions that shaped the company. He worries less about whether AI can produce a dazzling answer than whether the surrounding system can catch a quiet mistake.
The durable idea in the Invofox story is not really about invoices. It is about attention. The founders paid attention to the problem customers already had, to the segment pulling hardest, to the unglamorous failures hiding behind a demo, and to a partnership they wanted to preserve through every pivot. A messy page becomes useful data one field at a time. A company, it seems, is parsed much the same way.