HU-01Human
Platform pilot
You, or your data lead
Approves, signs, arbitrates. Has the final say.
signs
Patrimoine
Closed beta · installed in your cloud
Run by agents.
/po
/quality
/insights
Your data platform.
Run by agents.
Augmented by humans.
Patrimoine installs a complete, state-of-the-art data platform in your cloud and operates it day to day. Ten agents handle the run; a single person stays on call.
/po/quality/insightsThe product
The platform runs around the clock, even when nobody is there.
Around the clock, the agents load, test, watch and prepare fixes in the sandbox. During the day, your teams build with them, faster than ever before. And whenever you come back, everything is written down: what was handled, what awaits your signature, the KPIs to watch.
Waiting for you.
ReviewPR #142 · Make orders reliableQuality · 02:06
Review
SignatureLogidis migrationModelling · 02:04
ApproveQuestionWhy is margin out of bounds in Roubaix?Quality → Romain Carle
Follow up
GovernanceMargin target adjustmentawaiting Claire Vasseur
CertifyMaison Verdier demo: a fictional retailer, data generated around the clock, a real infrastructure running it. How the demo works
The team
Ten agents, and the humans who guide them.
Each agent takes over a data-team role, with its mission, its tools and its contract. You call them with a /, like a colleague. Everyone works around the clock; the humans approve, sign and give the context.
AG-01Agent
PO
Data product owner
Coordinates the team and writes the morning report.
/po
AG-02Agent
Ingestion
Data engineer · loads
Loads your sources around the clock, without losing a row.
/ingestion
AG-03Agent
Modelling
Data engineer · models
Builds and evolves your data models.
/modelling
AG-04Agent
Quality
Data quality
Tests every table before anyone reads it.
/quality
AG-05Agent
Operations
Data engineer · run
Watches the platform and contains incidents, 24/7.
/operations
AG-06Agent
Insights
Data analyst
Tracks your KPIs and flags what moves.
/insights
AG-07Agent
Catalogue
Data steward
Documents every table, definition and decision.
/catalogue
AG-08Agent
Costs
Data FinOps
Keeps the cloud bill under control.
/costs
AG-09Agent
Security
Ops and security · permissions
Checks access and protects personal data.
/security
AG-10Agent
Flows
Ops and security · outbound flows
Delivers data to your tools: ERP, CRM, AI assistants.
/flows
HU-02Human
Business and field teams
Finance, stores, supply…
Give the context and explain the numbers.
signs
HU-03Human
Leadership
Executive team, managers
Sets the company’s ambitions.
signs
The benefit
What a data team does, for a fraction of its cost.
A data team is often heavily tied up by the run and technical debt: keeping things running, fixing, watching, at the expense of building. Patrimoine gives that time back: the agents take the run, quality and daily reports; maintenance costs fall; the team focuses on what creates value, and time to market improves.
Source: Fivetran, The Enterprise Data Infrastructure Benchmark Report 2026 (March 2026, 500 data leaders at companies with over 5,000 employees): 53% of engineering capacity goes to maintaining and troubleshooting pipelines.
Incidents
Something breaks
at 2 am.
The Operations agent notices, contains it, tests a fix in the sandbox, and writes down what happened in plain language. You read it in the morning.
02:00
Detect
A broken data contract, a late flow, a saturated machine.
02:01
Protect
Hold the load, keep yesterday’s figures displayed as such, suspend outbound flows.
02:04
Test every action
Prepare and test the fix in the sandbox, never directly in production.
08:12
Report back
What happened, what was done, what was checked, what remains: everything is written and timestamped. The fix awaits your signature.
What stays human
Agents surface the facts.
Your teams give the context.
It is a virtuous loop: every piece of context your teams give is reread by the agents right away, and comes with the facts that follow. Your teams train the agents, day after day, and your know-how stays in the platform.
- 01The agents surface a fact
- 02Your teams give the context
- 03
The agents reread it right away - 04The next fact comes with its context
PR #142Make orders reliable
ReviewSign
Platform pilot
Approves, signs, arbitrates. A single person on call is enough.
Textile margin is down 2.1 pt.SMContext · Sophie MarekClearance extended by a week.
Business and field teams
They give the context, decide and sign their explanations. Patrimoine asks them: an invitation, never an order.
GROSS MARGINcertified v3
28.4%
Certified numbers
Every certified number carries a seal, a version, an owner and a one-line definition.
Examples from the Maison Verdier demo, a fictional company.
Explore
Everything an AI-native data team does.
Your data team now runs
around the clock.
Closed beta: we install Patrimoine in your cloud and work out the benefit with you, on your data.
qualityoperationssecurity