- Home
- Industries
- Legal and professional services
AI agents for law firms, notaries and professional services
Confidentiality is the first question in any firm. Our agents work from your documents, on a platform hosted in Germany, with the model route named in your contract. Lawyers, notaries and accountants stay the ones who advise.
The problems we hear first
Contracts against the playbook
Associates compare the same clauses with the same positions, deal after deal.
Data rooms
Hundreds of documents to sort and summarise before anyone can spot the issues that matter.
Intake and conflicts
New matters start with facts collected by email and a conflict search prepared by hand.
Client confidentiality questions
Clients ask where their documents go, who sees them and how long they are kept. The firm needs a precise answer.
Each one prepares. A person decides.
Examples of what we build in this sector, not client work. For each one: where a person approves, and what it never does.
Contract review against a playbook
Compares a contract with your positions, marks each deviation with the clause and the rule, and drafts the mark-up from your fallbacks.
- A person approves
- A lawyer accepts or rejects each finding.
- It never
- Advise the client or send the mark-up.
Due diligence assistant
Sorts and summarises the data room, extracts key terms into your template and flags what needs a lawyer, with the source page for each item.
- A person approves
- A lawyer reviews every flag before it reaches the report.
- It never
- Draw a legal conclusion.
Intake and conflict pre-check
Collects the facts, prepares the conflict search and the engagement file.
- A person approves
- A person opens the matter.
- It never
- Accept a client or clear a conflict.
Know-how answers for staff
Answers colleagues from your precedents and internal guidance, with the document it relied on.
- A person approves
- Partners own the guidance it reads.
- It never
- Answer clients directly.
One job, step by step, with the person inside the flow.
Due diligence data room
A data room for a share purchase arrives. The agent sorts and summarises it; a lawyer reviews every flag.
- Sector
- Legal
- Risk class
- Not listed in Annex III; reasons kept
- Who decides
- The responsible lawyer
- Data room received140 documents in 12 folders: contracts, leases, employment, IP (sample).Done
- Read inside the matter’s workspaceOnly this matter’s documents; nothing from other clients.GDPR Art. 5(1)(c)Done
- Key terms extractedChange of control, assignment, termination and exclusivity, into your template, each with its page.Done
- Seven change-of-control clausesEach may need consent for the deal; listed with the counterparty and the source page (sample).Flagged
- Red-flag summary draftedFacts and sources only, no legal conclusion.Drafted
- Waiting for the lawyerEvery flag is accepted or rejected by a person.AcceptRe-checkAdd to the reportAI Act Art. 14Waiting for a person
- Report section addedOnly after the lawyer’s review.Locked
- Run recordedEvery document opened, the findings and the lawyer’s decision on each.AI Act Art. 12On record
A sample, scripted on invented data: no model is called and nothing here is client work.
Where the law draws the line for you
Each rule with the day we read it. An engineer’s reading for your counsel to confirm, not legal advice. How we read the AI Act.
| Rule | What it says (our summary; quotes are verbatim) | What it means for an agent |
|---|---|---|
| AI Act Annex III 8Read 9 November 2026 | Lists AI used by or for judicial authorities to research and interpret facts and law, and in alternative dispute resolution. | A firm’s own review and drafting tools are usually not in this point; we write the reasons down. |
| GDPR Art. 28Read 9 November 2026 | A processor contract (DPA); sub-processors need the controller’s prior written authorisation, and the processor stays fully liable. | A DPA and a dated sub-processor list before any client document reaches the agent. |
| GDPR Art. 5(1)(e)Read 9 November 2026 | Personal data kept identifiable “for no longer than is necessary”. | Retention is set per matter in the contract; erasure comes with a receipt per step. |
| AI Act Art. 4Read 9 November 2026 | Providers and deployers take measures to support the AI literacy of their staff (as amended by the Digital Omnibus). | We brief the lawyers who review the agent on what it does well, where it fails and how to spot automation bias. |
| AI Act Art. 50(1)Read 12 November 2026 | Providers make sure people are told they are interacting with an AI system, unless that is obvious. Applies from 2 August 2026. | Customer-facing agents open with an AI notice in their first message. |
Nine layers around every agent, each with what is partial.
- Access and identityGoogle sign-in, revocable keys, an agent’s rights narrowed by its spec.Partial No MFA of our own and no enterprise SSO; open sign-up is on.
- Separation between clientsEvery record and search index filtered by the signed-in identity; a live agent gets its own containers and database.Partial Logical separation in one shared database, not yet tested independently.
- Hosting, network, encryptionHetzner, Nuremberg; HTTPS with HSTS; no inbound web ports; credentials sealed.Partial Stored data in general is not encrypted at rest.
- Only the data the task needsThe agent reaches only the systems its approved spec declares; secrets are scrubbed.Partial Personal data is not masked before the model.
- The model stepProvider set per deployment; each request admitted and recorded before it is sent.Partial The region is named in your contract, not promised EU-only.
- Untrusted content and testingDocuments are data, never instructions; a safety pack that must all pass.Partial No independent penetration test yet.
- A person decidesWrites wait for a person; a person approves every release; Stop and spend caps.Partial A live agent’s held actions are approved by our operator, not on your screen.
- Every run on recordEvery run recorded step by step, with approvals and who decided.Partial Append-only by convention, not tamper-evident; no client export screen.
- Deletion, export, backupsErasure with a receipt per step; a backup before every release.Partial Export and account deletion by hand, on request; backups not encrypted.
A timeline up front, a fixed price after one call.
Most teams start with a readiness review of the process they have in mind, or go straight to a pilot when the use case is clearly outside the high-risk list. We don’t publish prices: the proposal after a short scoping call gives a fixed price for your scope, in EUR, excluding VAT.
All engagements and what moves the priceBook a callPrice on request
AI Act and GDPR readiness for one use case.
What’s in itBook a callPrice on request
One agent on one process, tested and approved.
What’s in itQuestions from legal and professional services
Is an AI contract review tool high-risk under the AI Act?
For a firm’s internal review, usually not. Annex III point 8 covers AI used by or for judicial authorities, and in alternative dispute resolution, to research and interpret facts and law. A firm’s own review and drafting tools are not in that point; GDPR and professional secrecy set the tighter limits.
Is client data used to train models?
Not by us. The model provider and route for your agent, and their terms, are named per deployment in your contract.
Where are our documents kept?
On the platform in Nuremberg, Germany, in your firm’s own workspace, separated from other clients by the signed-in identity. The model step is named in your contract; we don’t promise that it stays in the EU unless that route does.
Can it run on our own infrastructure?
If your policy requires it, we discuss it in scoping. It changes the price and the timeline.
Related pages
Posts on these rules
Bring one process. Leave with a risk class, a pass bar and a plan.
A 45-minute scoping call with the people who would build it. You get a first view of the AI Act risk class, the controls it needs, and whether an agent is the right tool at all.