Your AI Lab. Owned Intelligence on your perimeter.
Forward-deployed engineers build a company AI lab: open-weight models fine-tuned or RAG’d on your use cases, on-prem or on a French sovereign cloud—then we run it for you. The outcome is Owned Intelligence: models, weights, data and prompts that stay yours.
Owned Intelligence — the outcome of the lab
Owned Intelligence is not a separate product. It is what the Company AI Lab delivers: local (or French-cloud) models specific to how your marketing, sales, operations and accounting actually work—with sovereignty, privacy and efficiency in one stack.
You buy the lab engagement. You keep the intelligence.
- Models on your infra or a French cloud—not a public generative API on raw client data.
- Fine-tune or RAG on your documents, jargon and workflows—not a generic chatbot.
- You own models, fine-tunes/weights, data and prompts end to end.
- Efficiency from compact specialized models (7–14B) and pipelines vs unbounded generic API spend.
This is your Company AI Lab on your perimeter—distinct from Zenaiyo’s Playbook Lab, which compounds our delivery playbooks and evals across engagements.
Sovereignty. Privacy. Efficiency.
Sovereignty
Sensitive work stays on infrastructure you control. Deployment target is chosen per client and checked on the current certificate.
Privacy
Professional secrecy, trade secrets and client files do not transit public generative APIs. Pseudonymisation and access control are built in.
Efficiency
Small specialized models and document pipelines process high volumes for a fraction of generic API cost—offline when the site requires it.
Owned Intelligence by function
Same lab. Models and workflows tuned to how your teams buy and operate.
Marketing
Brand-safe drafts and campaign follow-up on a private model over your docs—not a public chat on campaign data.
See AI agents →Sales
Lead qualification, research and follow-up with CRM context that never leaves your stack.
See AI agents →Operations
Document AI into ERP, tender RAG, bookings and triage—air-gapped when plants require it.
Industry Air-gapped Lab →Accounting
Invoice extraction and client-mail drafting on a local model—Secret-Pro path for cabinets.
Accounting sector →Offer ladder (prices HT)
Start with a short audit. The audit fee is credited in full against the Pilot if you continue within 60 days. Hardware and cloud pass through at cost.
Sovereignty Audit
€3,000–6,000
Data and workflow map, regulatory scan, top 3 use cases, build-vs-buy, infra choice, and a costed roadmap. Credited 100% against the Pilot within 60 days.
Pilot Lab
€15,000–35,000
One use case end-to-end: open-weight model on sovereign cloud or a GPU box, RAG and/or LoRA, eval vs a generic model, pseudonymisation gateway, governance pack lite, go/no-go report.
Lab Build with FDE
€60,000–180,000
2–4 use cases on one platform: gateway, model registry, fine-tune pipeline, RAG with access control, evals, observability, SSO, full governance pack, team training. Engineer embedded 2–3 days/week.
Managed Lab
€2,000–8,000 / month
Monitoring, model and security updates, quarterly re-fine-tune, eval reports, cost reporting, compliance evidence refresh, SLA. Infra pass-through on top.
Forward-deployed engineer à la carte: €800–1,000 HT / day (or 4–10-day blocks).
GPU cloud and on-prem servers are passed through at cost (or with a disclosed margin). Deployment target is chosen per client and checked on the current certificate.
What the lab contains
A production internal lab—not a chatbot demo—with governance you can show a DPO or security reviewer.
- Infra: GPU on-prem (1–2× L40S/RTX-class for 7–14B) or sovereign cloud tenant; network isolation; SSO.
- Model gateway: OpenAI-compatible internal API; sensitive prompts stay local; generic prompts can use an approved API; pseudonymisation layer.
- Models: open-weight base (Mistral Small / Ministral, Apache 2.0) plus LoRA fine-tunes; domain classifiers; OCR/document AI; speech-to-text.
- Knowledge: RAG over private docs with access control mirrored from your systems.
- Evals & observability: test sets per use case, before/after vs a generic model, tracing, cost per request.
- Governance pack: data map, DPA, AI Act literacy session, AI disclosure, model cards, evidence pack as needed.
- FDE: an engineer embedded 2–3 days/week who finds use cases, builds, and hands over.
We do not host your regulated data
Zenaiyo deploys into your environment or a certified host you choose. We are not HDS- or SecNumCloud-certified and we do not hold regulated data ourselves. Certifications are checked on the exact offer and date.
Sector entry offers
Three first offers for mid-market buyers who need sovereignty now.
Secret-Pro Lab
Audit + Pilot €18–30k, then Managed €2–4k/mo
Document AI into your accounting software plus client-mail drafting on a local model. Same budget many firms already spend on ChatGPT, Claude or Copilot—data stays in France.
Open sector page →Law firmsPseudonymisation gateway + firm RAG
Pilot €15–35k · Lab €60–150k
A local NER strips names before any cloud call. Then RAG over your precedents and templates. Built for the bar’s professional secrecy rule.
Open sector page →Industry & aero supply chainAir-gapped Lab
Audit + Pilot €25–40k + hardware
Domain model on standards and part numbers, document AI into the ERP, on an on-prem GPU box. For trade secrets and export-control data.
Open sector page →Frequently asked questions
What is Owned Intelligence?
The outcome of the Company AI Lab: models, fine-tunes/weights, data and prompts that stay yours and are specific to your use cases. It is not a separate SKU you buy beside the lab.
What is a Sovereign / Company AI Lab?
Your lab on your perimeter: open-weight models, RAG or fine-tunes on private data, a hybrid gateway, evals, and a governance pack—built and run with a forward-deployed engineer. Distinct from Zenaiyo’s Playbook Lab, which compounds our delivery playbooks.
Do you host our client or health data?
No. We deploy into your environment or a certified host you choose. Zenaiyo is not HDS- or SecNumCloud-certified and does not hold regulated data.
Who owns the models, weights and data?
You do. That ownership is Owned Intelligence—the closing promise of the engagement.
How does the Sovereignty Audit relate to the Pilot?
The audit is €3–6k HT over 3–5 days. If you start a Pilot within 60 days, the audit fee is credited in full against the Pilot price.
What does Managed Lab include?
Monitoring, model and security updates, quarterly re-fine-tune, eval reports, cost reporting, compliance evidence refresh, and an SLA—plus infra pass-through. Typical range: €2–8k HT per month.
Start with a Sovereignty Audit
Three to five days. A costed roadmap you keep. Credited in full against the Pilot if you continue within 60 days.
Book a Sovereignty Audit