became ai
as the impossible became ai, reality followed
We use AI to solve problems that were considered unsolvable until recently: fast, at a fixed price and at the state of the art. We come from AI research at TU Wien, and we build systems that hold up in daily operation.
Do you have a problem nobody can tell you is solvable? Those are the ones we are interested in. Contact →
What we do
AI Sprint: from idea to working prototype in three to six months
At a fixed price agreed in advance. You get a system that works, a reliable answer to the question of feasibility and an honest estimate of what it will take to put it into operation. If the problem convinces us, we carry part of the risk ourselves.
Into operation
We run what we build, on your own servers if you wish. Then your data never leaves the building.
Orientation
Finding, assessing and prioritising use cases. We also tell you which ideas are not worth pursuing. We don’t sell projects we don’t believe in ourselves.
Training
Drawn from our teaching at TU Wien and tailored to your organisation: understanding AI well enough to take part in the discussion for executives, hands-on skills for specialist and development teams, up to agentic software development.
Confidential sparring for leadership
For board members, managing directors and senior civil servants we offer one-to-one conversations without an audience. Ask every question about agentic AI, governance and liability that you would rather not ask in front of your own organisation.
Applied research
Research collaborations and funding applications, together with partners from universities and research institutions.
Why people come to us
“We have hundreds of thousands of pages nobody can read.”
Handwriting, scans, forms, legacy holdings. We make them searchable and index their content, with review workflows so that experts stay in control.
“When Ms M. retires, her knowledge goes with her.”
In the coming years many experienced colleagues will retire, taking with them knowledge that was never written down. We help secure that knowledge while it is still there, and make findable what lies scattered across files and documents.
“Our people use ChatGPT, including with data that doesn’t belong there.”
Models running in-house and anonymisation before any external processing: an official channel that is better than the unofficial one.
“We search by keyword and still don’t find what belongs together.”
CV and vacancy, start-up and corporation, a current project and comparable past ones. We match what fits in meaning, even when the terms don’t.
“We have twenty AI ideas and don’t know which one is worth it.”
We assess them with you, build the most promising one in an AI Sprint and tell you openly which ones aren’t worth it.
“The board is supposed to decide on AI, but nobody wants to admit what they don’t know.”
That is what our confidential sparring is for: one to one, without an audience.
And often it is something that doesn’t have a name yet.
Selected work
- Vienna City LibraryHistorical handwriting, machine-readable
- A vision language model of our own, trained on the European supercomputer Leonardo. Four models vote word by word and, on unseen pages of Kurrent, the historical German handwriting, reach a quality that opens up extensive collections. The AI does the reading, the library’s experts make the decisions.
- International retail groupThe entire project portfolio in view
- A large portfolio of digital projects, centrally managed but never looked at as a whole. We make it visible in its entirety for the first time: which projects resemble each other, which past projects are similar to a current one, and how did they turn out? That way, projects at risk show up while there is still time to steer.
- Financial services providerOfficial documents, processed automatically
- For an Austrian financial services provider we process official documents from PDF all the way to matching against the company database. The pipeline runs in daily operation on the client’s own AI server. No document goes to a third-party cloud.
- National trust labelIT security
- We built and operate the platform: the application process for three levels, the connection of audit partners and the automatic evaluation of the security questionnaires. In use for years.
- Industrial groupThe system the business relies on every day
- For an Austrian industrial group we built the central planning and staffing system: time tracking, deployment and project planning, apprentice qualifications, interfaces to payroll and other systems. We have been developing it further for years.
We are happy to name further references in person.
Why not simply a big consultancy?
The expertise is at the table
There is no pyramid of partners and junior consultants. The people you talk to also work on your project.
We know the state of research because we help shape it
We teach generative AI at TU Wien, and Georg Gottlob has been researching artificial intelligence for decades. Teaching means knowing what is becoming possible.
Results, not slide decks
After three to six months something is running. Or you know, with reasons, why it can’t be done. At a fixed price.
Independent
We have no partner agreements with cloud or model providers and no licences to sell. We recommend what works for you. Often that is a model running in your own building.
Research that runs in practice
People expect papers from universities. Our systems run in daily operation, for clients who rely on them every day.
Focus areas
- Understanding documents. Forms, official documents, handwriting, historical holdings: exactly where conventional text recognition fails.
- Preserving knowledge. Securing experience and making knowledge scattered across files and documents findable.
- Assessing texts. Judging against a transparent standard: job descriptions, submissions, contracts.
- Matching and recommendation. Bringing supply and demand together by meaning.
- Local, sovereign AI. Models that run on your premises and are trained on European infrastructure.
- Agentic systems. How to actually build software with AI. We do it every day and teach it.
- AI Act and governance. What you need to do, and how to demonstrate it credibly.
- Data protection through semantic anonymisation. Measuring re-identifiability and reducing it in a targeted way.
- Bilateral AI. Combining neural networks with rules so that decisions remain verifiable.
Stance
AI is a skill, not an actor
We build systems that strengthen people in their work and leave the decision with them.
Not every problem needs a chatbot
A chat window shifts the work onto the user: they have to know what to ask and check every answer. We build AI into the place where the work already happens, with results that can be checked at a glance. And agents get exactly as much room to act as the task requires.
Sovereign intelligence
Being confident in dealing with AI, and remaining sovereign over your own AI. For organisations, that means their own competence. For Austria and Europe, it also means their own infrastructure and control over their own data.
You need to see why the AI does what it does
Every assessment needs its reasoning, every recognised text its source, every recommendation the cases it is based on. That is the only way to keep a system’s decisions verifiable.
Digital humanism
We are committed to the Vienna Manifesto on Digital Humanism and serve on the board of the Digital Humanism association. AI should widen people’s room for action, not narrow it.
For policy makers, the public sector and the media
Austria is making decisions about artificial intelligence right now that will have effects for decades: in public administration, education, research and infrastructure. These decisions deserve first-hand advice.
We are available to ministries, public authorities, interest groups and newsrooms for background talks, statements, expert opinions and lectures. We are independent of vendor interests and know what research can do today.
AI in public administration
Many public offices have long been using AI, often through private accounts with external services and sometimes with data that should not go there. At the same time, federal and state governments are building central services. We offer an independent assessment: which tasks are suitable for AI, what the available tools achieve there and what to look out for with data and data protection. On request, we accompany a small pilot with criteria agreed in advance for when it ends.
Topics we speak on: AI in public administration · sovereign AI infrastructure for Austria · AI in schools and universities · implementing the AI Act · AI and public discourse · digital humanism
Teaching, research, talks
Generative AI at TU Wien
We have taught the subject since 2023, now in two courses (theory and practice), with well over a thousand students and hundreds of supervised team projects so far. Consulting firms also send their staff to our courses. Knowledge management is a recurring focus of these projects. We work closely with the Christian Doppler Laboratory for Recommender Systems.
Research
Patents on information extraction, access documents and conversational decision processes, publications on visual document understanding, supervision of master’s theses.
From our research: semantic anonymisation
Anyone who wants to share documents containing personal data, or process them with AI, has to anonymise them. Redacting names is not enough: occupation, age and neighbourhood together make a person recognisable. Our multi-agent approach finds exactly these indirect identifiers considerably more reliably than conventional methods and removes them from the texts.
Talks
We speak regularly at international conferences, at universities, in continuing education and in companies, including the Wiener Vorlesung, the DigHum Lecture Series and the TU Wien Academy for Continuing Education.
Coming up
5 November 2026 · AI:Advanced, TU Wien Academy for Continuing Education: Bernhard Krüpl-Sypien speaks on generative AI.
Team
Georg Gottlob
First AI professor at TU Wien, professor at Oxford and Calabria. Fellow of the Royal Society, Wittgenstein Award, Ada Lovelace Medal. Serial founder, each company successfully acquired (McKinsey, Meltwater).
Bernhard Krüpl-Sypien
Managing Director. Teaches generative AI at TU Wien and coordinates the Generative AI group at CAIML. Software architect, inventor of several patents, board member of the Digital Humanism association.
Johannes Oster
Lead AI Engineer. Teaches generative AI at TU Wien and speaks at the Wiener Vorlesung and in the DigHum Lecture Series, among others.
David R. Prasser
Entrepreneur and innovation strategist, founder of StartUs Insights. He brings the market’s perspective: which ideas stand a chance, who else is already working on them and how a prototype becomes a business.
Contact
Tell us what you are stuck on.
A first conversation is free and without obligation. Afterwards you will know whether an AI Sprint is worth it.
hello at became.ai
(Apply your human knowledge to turn this into an email address.)