AI that knows your matters

The tools will change. Your knowledge shouldn't have to move with them

The legal AI market is moving too fast to lock your business into one vendor today. What lasts is what you know yourselves: your contracts and matters, your judgement, and your experience of what actually held up.

In practice, an AI solution has three parts: the interface the lawyer works in, the engine that performs the task, and the knowledge it works from. The first two can often be bought, and several of them are very good. If you control the engine, and where and when it can be used, you can also start integrating legal work into other processes and systems across the business. The third no vendor can sell you, because it only exists with you.

We build it together with your lawyers, and we build it so you can take it with you. That way you can change model or vendor in two years without rebuilding the business.

Knowledge, not a connector

Connecting an AI service to your document management system or contract database is not the same as giving it your knowledge. Let a model loose on twenty years of documents and it will find a lot. But how would it know which contract was a good one? Which clause held up in negotiation and which was struck out? Where you usually give ground, and where you never do?

Your lawyers know. We sit down with them and structure that knowledge: what is your established practice or contract standard, what has been reviewed, and what has been cleaned of sensitive information and can therefore be used widely. It isn't the most spectacular part of an AI project, but it's where quality is decided.

‍

Buy, build or both

‍The question is rarely which tool is best, but which one fits your caseload, your ways of working and the systems you already have. For an in-house team of around ten people, it isn't obvious which system to use, or how. Sometimes a general-purpose model does the same job, if it's built on your knowledge and runs in the systems the company already uses. For a firm with hundreds of lawyers, the calculation can look completely different. When you buy, we help you set the requirements that matter in the long run: where data is stored, what the vendor may use it for, and how you exit the contract with your material intact.

Knowledge architecture

‍A model can read everything you have, but it can't judge what was good. Your lawyers do that. We turn what they know into something AI can work from, whichever tool sits on top.

Agents and workflows in your systems

‍We work as legal engineers, with lawyers and developers building together. We build agents and workflows in the tools you already use, your AI platform, Word, Teams and case management systems, based on how your matters actually run.

When the documents in a matter are complete, a workflow can, where appropriate, fill in your templates and place a first draft with the responsible lawyer. When the business sends in a contract, it can be reviewed against your playbook and come back with the deviations marked. The lawyer gets the work, not yet another tool to remember. The workflow is built in or outside the tool, and the knowledge is owned by you.

Measurement that holds up when someone asks

‍How many people log in says little about what AI does for the business. We measure time per matter type before and after, how much of the drafts pass review, and which errors are caught. We agree the measurement points before we build.

At a law firm, that's the evidence you need when a client asks what AI means for your fees. In an in-house team, it's the evidence for what can be handled internally instead of bought externally, and for how quickly the business gets answers.

We also work with classic machine learning where it does more good than a language model: classification, anomalies and forecasting.

Adoption and training

We train on your own material and your own matters, not on generic prompt technique. The most important skill is being able to review: seeing where an answer comes from, and recognising when the source doesn't support the claim.

Technologies and systems we work with

We build agents where the system answers based on your own documents, contracts, matters and the knowledge in your lawyers' heads, with citations back to the source document, so the answer can be checked. We build integrations with case management, document management and time recording systems, so that AI works with the material you already have rather than alongside it.

When you buy off the shelf, we help you evaluate the market and set the right requirements on where data is stored, what the vendor may use it for, and how you exit the contract with your material intact.

We are vendor-independent. We recommend what actually solves the problem, not what happens to be closest at hand.

How we know it works

Digitala Juristen. A public AI service in live operation at Familjens Jurist, developed with Yellow Elk as technical partner and part-funded by Vinnova, Sweden's innovation agency. It answers questions from the public on topics such as custody and divorce. The firm's legal knowledge was built as a separate layer, so the same knowledge can be used in the lawyers' own tools.

[Read more about the project at Familjens Jurist →]

Lawyers and developers in the same team. Most AI projects don't stall on the technology, but because the domain experts aren't in the room. The hub is led by Daniel Lönnberg and Martin Wass. Daniel is a lawyer and was previously Head of Data & Innovation at Familjens Jurist, where he led the AI work. Martin is an AI and solutions architect who has led several cross-functional teams over the years. The team includes lawyers who have introduced AI in working law firms themselves, as well as solutions architects, data engineers and legal engineers.

A data company behind us. Law & Technology is part of Yellow Elk, a Nordic data and AI consultancy with more than 200 consultants. An in-house legal team sits right in the middle of the company's IT environment, and building data platforms and AI in exactly that kind of environment is Yellow Elk's core business. We're not an AI company that learned data. We're a data company that built AI.

We stand behind it in public. We speak about how firms stay in control while scaling with AI, among others at TechHeads in Kalmar with the talk "Don't outsource your brain – how law firms stay in control while scaling with AI", and at TechTorget, where we talked about why you should build the knowledge layer yourselves and buy what's already good.

[TechTorget Copenhagen →]

[Read more about TechHeads Kalmar →]

‍

Tools that were bought but never connected to the firm's own material. Uncertainty about what professional secrecy allows, so nothing gets done. Vendor contracts where it's unclear who owns the data and what happens if you want to switch. In-house teams choosing between ten tools with no way to compare them.

And most often: a firm that tried something, saw it work in the demo, and then found that nobody uses it. The tool didn't know the matters. Nobody had decided who owned the rollout. And a lawyer who can't check an answer won't trust it, which is entirely reasonable.

That's where we make the biggest difference, when AI becomes something lawyers can trust and use in their daily work.

Let's talk data. We'll help you find the right way forward, wherever you are today. ContactDaniel Lönnberg or Martin Wass
‍