The Luxembourg founder who turned €2.8mn into a bet on legal AI

Luxembourg is one of Europe's most concentrated legal markets. Its economy is built around finance, funds, holding companies, cross-border structuring and European regulation — sectors in which legal advice functions less as a support service than as part of the infrastructure. Lawyers work across languages, jurisdictions and regulatory regimes, often for clients whose headquarters, assets or obligations stretch well beyond the country's borders.

It is in this environment that Tim Kerger, a young Luxembourgish founder, has built Clerk, an AI legal start-up that has raised €2.8mn in funding at a €6.5mn valuation. Trained in economics and shaped by work in cybersecurity, he approaches legal AI less as a spectacle of automation than as a problem of systems, incentives and trust. Economics informs how he reads the market he is entering; cybersecurity, how he thinks about the risks lawyers cannot afford to ignore.

Tim Kerger describes what he believes will decide the field in plain terms:

“Legal AI will not be won by the best chat window. It will be won by systems that can actually work.”

Beyond the chatbot

That distinction sits at the heart of Clerk. Rather than positioning itself as another wrapper around a large language model — a box for lawyers to type questions into and hope the answer is usable — the company describes Clerk as an agentic legal workspace: an environment in which an AI agent can move through a legal matter much as a junior associate might.

According to Clerk, the agent can open a file, read documents, search jurisprudence, identify applicable law, calculate deadlines and indemnities, draft legal correspondence, prepare conclusions and export documents on a firm's letterhead. The design intent is not simply to respond to prompts, but to give the system tools, context and a workspace in which to act.

The company's technical bet is unusually specific. Clerk gives the AI agent access to a specialised Luxembourg jurisprudence database, built so that case law can be found through natural language rather than narrow keyword search. The corpus has been embedded using 3,000-dimensional vectors, which the company says lets the system map legal meaning rather than merely match terms. A lawyer can ask a question in the language of a matter — about abusive dismissal, non-compete clauses, proportionality, procedural deadlines or damages — and the agent searches for decisions that are conceptually relevant, not only those containing the exact same words.

For lawyers, that difference could matter. Legal research is rarely a matter of finding a single phrase; it is a matter of finding the line of reasoning, the pattern of facts and the judicial treatment that apply to the case at hand. A keyword search can miss a useful precedent, and a general-purpose chatbot can invent one. Clerk's stated aim is more disciplined: to search the Luxembourg corpus, open the decisions that matter, read them and cite them back to their public source.

Built for Luxembourg first

In Luxembourg, the case for such a product is relatively strong. The market is sophisticated, multilingual and internationally exposed, and its lawyers often operate at the intersection of local law and European structures. Yet legal technology is frequently designed first for larger jurisdictions and adapted elsewhere, which can leave a smaller market treated as an edge case.

Clerk sets out to reverse that order. The platform starts from Luxembourg law and builds outward: its jurisprudence research is centred on Luxembourg decisions, its workflows are meant to reflect local practice, its drafting functions are designed around documents lawyers actually use, and its calculators address the deadlines, notice periods and indemnities that arise in real matters. The result, the company argues, is less a generic productivity assistant than a tool built for a specific jurisdiction.

Its broader claim is that the agent can work across an entire matter — researching jurisprudence and applicable law, reading documents, extracting facts, calculating amounts, producing drafts, preparing arguments, identifying weaknesses, collaborating with colleagues and keeping the file organised: much of what surrounds a lawyer's final judgment.

That breadth is what Tim Kerger sees as distinguishing Clerk from the first wave of legal AI tools. A chatbot answers a question; a workspace, in his framing, becomes part of the workflow. The distinction is also commercial: lawyers may be reluctant to pay much for another window onto a model they can already access elsewhere, but might pay for a system that saves hours inside a matter and makes the work more traceable.

Trust and confidentiality

Traceability is central to the pitch, because legal AI has a trust problem. Large language models are powerful but can be opaque, overconfident and wrong, and in law a fluent mistake can be more than embarrassing — it can be professionally dangerous. Clerk's answer, the company says, is to ground the agent in sources, show its reasoning and flag what cannot be supported, so that a weak link is visible before it reaches a client or a hearing.

Confidentiality is the other constraint. For legal AI, privacy is less a feature than a precondition for use at all. Clerk describes itself as built around anonymisation, EU hosting and the principle that identifiable client information should not be sent casually into third-party systems — a point that, in a profession shaped by secrecy and responsibility, may weigh as heavily as the model itself.

The €2.8mn round gives Clerk room to develop that infrastructure: capital for engineering, corpus development, security, product design and the slower work of selling into a profession that does not adopt new tools simply because they are fashionable. Curiosity about AI is widespread among lawyers, but curiosity is not the same as trust, and the company will have to earn its place matter by matter.

A different kind of leverage

Tim Kerger's wider thesis is that the next generation of professional software will be less passive — not merely storing documents and displaying tasks, but containing agents that can take instructions, use tools, consult specialised databases, produce drafts and return with reasoned work product.

For the legal profession, that would be a meaningful shift. The traditional firm is organised around human leverage: partners, associates, trainees and support staff passing work down the chain. Clerk proposes a different form of leverage, in which an AI agent handles a first pass of a matter while the lawyer remains responsible for strategy, judgment and client advice. In Luxembourg, where firms handle complex, high-value work without the scale of the largest legal markets, a small team using such a workspace could in principle research more deeply, draft faster and manage more matters without simply adding hours.

Tim Kerger is measured about what that means. The pitch is not that AI will make lawyers obsolete, but that much of the work surrounding legal judgment can increasingly be prepared by machine — so that the lawyer arrives at the centre of a matter sooner, rather than disappearing from it.

Whether Clerk can deliver on that at scale is an open question. Legal AI is a crowded field, and the profession is right to be cautious about tools that can confuse fluency with accuracy. But Clerk has chosen a clear position: local rather than generic, agentic rather than conversational, and built around the work of a matter rather than the interface of a prompt. For a market that has rarely produced legal-technology companies of its own, the round is a sign that investors see room for Luxembourg to build one.