Managing partner, knowledge-management partner, or COO/CIO in firms of roughly 30+ lawyers.

Law firms · France

A lab that respects attorney-client secrecy

Q: What is the first product?

A local NER that strips names before any cloud API call, then RAG over your precedents and templates. Pilot typically €15–35k HT.

Q: Do you sell to notaries the same chat tool as the national programme?

No. The profession already chose a national stack. We sell custom workflows on top—deed checks, document extraction, file intake—for large offices that respect the charter.

Q: Where does the model run?

On-prem or private EU cloud, with open-weight models. We do not send raw client data to a public generative API.

Client and case data never go to generative AI unless pseudonymised.

The bar’s 2026 ethics guide is clear: do not communicate client or case data to generative AIs without pseudonymisation or anonymisation. We sell a local pseudonymisation gateway first, then RAG over your precedents and a drafting model in house style—on-prem or private EU cloud.

Book a Sovereignty Audit All sectors

Pseudonymisation gateway + firm RAG

Pilot €15–35k HT · Lab €60–150k for mid-size firms

Buyer: managing partner, knowledge-management partner, COO/CIO in firms of 30+ lawyers.
Deployment: on-prem or private EU cloud; open-weight Mistral models.
Requirement: lawyers’ professional secrecy (art. 66-5). The CNB 2026 guide effectively bans raw client data in generative AI.
Notaries (secondary) — The profession already has a national Mistral + Scaleway programme. We do not sell the chat tool. We sell custom workflows on top—deed checks, document extraction, client file intake—for large offices that respect the charter.

See the Sovereign AI Lab ladder →

What firms buy first

1
Pseudonymisation gateway
A local NER strips names before any call to a cloud API. Sharp, cheap first product.
2
RAG over firm know-how
Precedents, templates and procedures with access control mirrored from your DMS.
3
Contract review classifiers
Clauses and risk flags for associate review—every output verified by a lawyer.
4
On-prem speech-to-text for meetings
Client meetings transcribed on infrastructure you control.

How Zenaiyo works

One flywheel: embed, ship agents, compound in our Playbook Lab, leave you owning the assets—GDPR-native.

Embed (FDE)

Forward-deployed engineers map how the business really runs—tools, handoffs, and where agents can act alone vs. where humans stay in the loop.

Ship AI agents

We wire agents into your stack with guardrails, logging, and handoff paths—bookings, follow-up, billing, admin and reporting.

Playbook Lab

Every deployment feeds evals and reusable playbooks. That internal lab compounds Zenaiyo’s delivery—distinct from your Company AI Lab.

Owned Intelligence

You own models, data, weights and prompts. That is Owned Intelligence. Need secrecy or French cloud? Start with a Sovereignty Audit.

Book a free automation audit Sovereignty Audit

Frequently Asked Questions

Sovereign AI for law firms — bar rules, the gateway, and pricing.

What does the CNB 2026 guide require?

Client or case data must not be sent to generative AI without pseudonymisation or anonymisation. Outputs must be verified. That is why we lead with a local pseudonymisation gateway.

What is the first product?

Do you sell to notaries the same chat tool as the national programme?

Where does the model run?

Who is the buyer?

Find the right workflow before you automate.

Use the free 10-question checklist to identify the workflow leak with the clearest time and ROI upside.

Run the Free Audit Checklist

No email required — takes about 3 minutes