---
title: "AI governance: rules, registers, evidence"
description: "A register of agents and models, rules of use and an audit trail of every action, with the evidence required by the AI Act, DORA and GDPR."
url: https://www.joinx.it/en/technology/governance
---

# The answers for the auditor, ready before the question

Faced with an agent in production, the auditor asks who answers for it, which rule it applied and who approved the decision. With Governance we record agents, models and actions as the work happens, so each answer is already written and dated.

Governance · Run log

-   09:12 · Draft quotation from a requestQuotation agent

    Completed
-   09:40 · Corrective action on a non-conformityQuality department

    Awaiting approval
-   10:05 · Check on a supplier invoiceSupplier invoice agent

    Being checked
-   10:31 · Answer on a technical data sheetTechnical assistant

    Completed

Illustrative interface

1.  [01Data Foundation](https://www.joinx.it/en/technology/data-foundation)
2.  [02Data Intelligence](https://www.joinx.it/en/technology/data-intelligence)
3.  [03Company Brain](https://www.joinx.it/en/technology/company-brain)
4.  [04Agents](https://www.joinx.it/en/technology/agents)
5.  [05Workspace](https://www.joinx.it/en/technology/workspace)
6.  [06Governance](https://www.joinx.it/en/technology/governance)
7.  [07Security](https://www.joinx.it/en/technology/security)

## What it contains

-   01

    ### Register of agents and models

    Lists the agents, models and applications in use, each with an owner, a purpose and a level of autonomy, and we update it as we build and modify the agents

-   02

    ### Rules of use and roles

    For each agent we write down the permitted uses, the data allowed, who answers for it and who approves its actions

-   03

    ### Run log

    For each action we record when it took place, on which data, under which rules and who approved it, in logs an auditor can consult and verify

-   04

    ### Evidence for audits

    We collect and update the project documentation for the AI Act, DORA and GDPR while the project is under way

-   05

    ### Test environments and synthetic data

    We test new agents and models in safe environments and with synthetic data, with no effect on customers

-   06

    ### Measuring value

    We measure the quality, cost and value of the new process against the starting point, and use the corrections and reports from the people who work on it to improve it

## The inventory of AI systems already in use

An email assistant, a service that summarises meeting minutes: this is what is known as shadow AI, tools already in use but outside any control. With Mapping we find it in the interviews, where people describe the tools they really use, and by cross-checking what emerges against the company's list of systems and licences. Then we enter it in the register, each tool with an owner and a risk level.

## How regulations enter the project

We classify each system under the AI Act before development, keep in the register the information on ICT providers that DORA requires and, for GDPR, apply data minimisation together with those responsible for data protection.

## Frequently asked questions

### Does it also apply to tools you did not build?

Yes. We also enter in the register the AI systems of most other suppliers, and connecting them takes some work to adapt them to your organisation. Those we do not connect stay in the register marked as not connected, so you know where you have automatic evidence and where you do not.

### Does it slow the work down?

Less than people fear, because we collect the documentation while the project is under way. Hundreds of projects at Join Business have taught us that the time given to this work during the project pays the organisation back later too, when an audit asks for the evidence.

### What happens when a model changes?

We treat it as a change to the process: an owner authorises it after the tests on the test bench, and the register keeps the date, the outcome and the rollback procedure.

### How does it differ from Security?

Governance sets the rules of use and keeps the evidence; Security enforces them on the systems and protects data, access and agents.

## Take stock of the AI systems already in use

Often the first step is knowing what is there, and we can start from that together with your control functions.

[Contact us](https://www.joinx.it/en/contact?tema=governance)
