An AI internal knowledge base agent makes the firm’s past work searchable: document templates, vetted clauses, research notes, memos. When a lawyer asks something, they get the relevant internal material, with its source, and a flag if the statute it cites has been amended since. Access rights per case file stay as the partners set them.


The most valuable thing in a law firm, after its people, is the work already done. A research note on a limitation period issue, a limitation of liability clause negotiated ten times, a template claim that has made it past the court’s filter many times. They usually sit in folders per case file, in emails and on one associate’s computer. Anyone who does not know they exist rebuilds them from scratch.
Indexes the document templates, contracts, notes and memos in the firm’s archive and makes them searchable by legal issue, not by file name
Links each piece of material to the statutes and decisions it relies on and checks in Lege5 or Sintact whether they have been amended since
Flags old templates, for example those written before a legislative change, so they are not used without review
Builds a clause library, with the variants used and the context in which each was accepted
Answers “have we worked on this before” questions, with a pointer to the exact document and the lawyer who wrote it
The answer is not an opinion. It is the relevant internal material, with its date and a flag for whatever has changed since.
The knowledge base also covers what is not a document as such. A partner’s observations on how a particular court tends to rule, the lessons from a lost case, the reason a clause was rejected by an important client. When lawyers choose to write these down, the AI agent links them to the practice area and the type of case they relate to, so they surface when someone works on something similar.
Tell us what slows you down. We talk, then decide together whether and how we work.

On Monday, an associate gets a case on a director’s liability in an insolvency. They ask the AI agent whether the firm has worked on this before. They get two research notes from recent years, a claim filed in a similar case and the name of the partner who led it. The AI agent adds that an article cited in one of the notes has been amended since.
On Wednesday, a partner finishes a new memo. The AI agent suggests where it should go in the knowledge base, which tags it should get and which older materials it should be linked to. The partner confirms, and the memo becomes searchable for those with access to that practice area.
On Friday, the person in charge of the knowledge base gets the list of materials flagged as possibly outdated that week, for example templates that cite a repealed provision. They decide for each one: update it, archive it or keep it with a note.
A trainee finds the firm’s templates and the notes on recurring topics on their own, without interrupting a partner at every step. For anything not in the knowledge base, the AI agent tells them who in the firm worked on the subject, and the question is logged, so it is clear where material is missing.

Google Drive, SharePoint or OneDrive, wherever the firm’s document archives live
Practice management software such as Juridis, IMFS, AvocatPRO or EBAS Iuris, for the link between documents and case files
Lege5 and Sintact, to check the version in force of the statutes cited and for reference case law
The firm’s email, only for the mailboxes and folders the partners choose
Notion or Airtable, if the firm already keeps internal notes and clause lists there
Because a law firm’s archive is covered by professional secrecy, here we almost always recommend a harness that runs on the firm’s own infrastructure. The documents do not need to leave the firm to become searchable.

It does not give legal opinions or decide which clause variant suits a client
It does not get around the access rights set per case file and per client
It does not draft final documents from the knowledge base, it brings the materials the lawyer starts from
It does not send internal documents outside the firm
Nor does it treat an old note as current truth. Every piece of material comes with its date, and when the statute it relies on has changed, the AI agent says so before the note is used. And it does not fill out a note with case law it did not actually find in the source.

You start with one practice area, usually the one with the most repeated material, and a single archive folder. We first discuss where the AI agent runs and how the firm’s access rights are reproduced, because here the architecture is decided before anything else.
The AI agent indexes the material in the chosen area and takes a first inventory: which templates exist, which are duplicates, which cite amended provisions. A partner reviews it and confirms the official templates. After a few weeks in which the team has used the knowledge base for that area, you add the next one. You see the cost of the storage and search tools before signing.
The partners decide what goes into the knowledge base and who sees what. Access follows the firm’s rules per case file and per client, including the separations needed to avoid conflicts of interest. A lawyer who does not work for a client does not get that client’s documents in search results, even when they are technically relevant.
A person also confirms what becomes an official firm template. The AI agent can suggest that a frequently used clause go into the library, but validating it is a professional decision. And any material from the knowledge base used in a new document is checked by the lawyer who signs it, because responsibility for the final document is theirs, not the knowledge base’s.
The most important link on the page. From here you go up to the agent we sell.
Sideways, on the same level, so you can move easily between pages.
They do not need to. For a law firm’s knowledge base we almost always recommend a harness that runs on the firm’s own infrastructure, so the documents stay where they are now. We have that conversation before signing. If your confidentiality requirements cannot be met technically, we tell you from the start.
The AI agent takes on the firm’s access rules: who works for which client and on which case file. Search returns only the materials the person asking has access to. The separations for conflicts of interest are set by the partners and tested during configuration, on real cases, before the knowledge base is opened to the team.
The AI agent periodically checks, in Lege5 or Sintact, the statutes cited in the templates and notes in the knowledge base. When one has been amended, it flags the material and notifies the person in charge of that area. Until it is updated, anyone who finds the template in search sees the flag and the date of the amendment, so they do not use it without review.
It can bring the right templates, the clauses used before in similar situations and the research notes linked to them, so the lawyer does not start from a blank page. It can also prepare a working outline. The final drafting, the choice of clauses and responsibility for the document remain entirely with the lawyer who signs it.
The retention rules are set by the partners, in line with professional obligations and with what was agreed with the client. The AI agent applies the rules: it can remove that client’s documents from search, keep only anonymized templates if that is what you decided, and keep a record of what was done. The decision on what is kept remains with the lawyers.
Tell us what slows you down. We talk, then decide together whether and how we work.