---
title: "AI governance in a company: policy, roles, accountability · Volcanic Minds"
description: "How to govern AI in an established company: list of uses, who approves, which data, how to measure, who is accountable and what the law requires."
url: "https://volcanicminds.com/en/guides/ai-governance"
lang: "en"
type: "second_level_page"
updated: "2026-10-31"
alternate: "https://volcanicminds.com/guides/ai-governance"
summary: "You start from a list of what is already running, use by use, and for each one you decide whether to continue, correct or stop it. Then you need one person who coordinates, with a written mandate from management, and a small group with IT, legal, data protection and HR; data rules split into classes; approval proportionate to risk, with a fast lane for simple cases; a number to measure, chosen before you start. When AI gets it wrong, three levels are accountable: the company towards customers and third parties, the process owner inside it, and whoever uses the tool, who must check the result. The AI Act does not require a private company to have a committee or a list of its systems: they are the tools that let you prove you meet what it does require."
---

# Governing AI in an established company who decides, which data, who is accountable

For whoever has to bring order to the AI tools that departments already use: who approves a new use, which data can go into it, how to measure it, who is accountable when it gets things wrong and what the law really requires.

## Where do you start governing AI in a company?

From a list of what is already running. In an established company AI arrives before the rules: in 2025, 71% of large Italian companies had at least one AI project, against 59% in 2024, and 84% had bought licences for at least one generative AI tool, led by Microsoft Copilot, ChatGPT Plus and Gemini Advanced. The rules lag behind: four large companies in ten have usage guidelines, and only 9% have structured AI governance, “with defined responsibilities” ([Artificial Intelligence Observatory of Politecnico di Milano, February 2026](https://www.osservatori.net/comunicato/artificial-intelligence/intelligenza-artificiale-italia/)).

Asked by an operations manager at a 600-person company, with ChatGPT, Copilot, a chatbot and two prototypes already in use, Google and the AI assistants give the same recipe: a committee, one person who coordinates, a list of systems, traffic-light data rules, approval by levels. They disagree on who is accountable when AI gets it wrong, and almost none says what Italian law requires: of the 73 pages we read, across Google results and the sources of the AI answers, Law 132 of 2025 appears in 7 and Legislative Decree 179 of 2026 in none.

The list goes by use, not by tool. ChatGPT, asked by someone working in compliance, advises recording “the system and the use case separately”, and it is the right advice: the same Copilot that summarises a meeting can write the reply to a complaint, and the two uses carry different risks.

| Field | What to write | What it is for |
| --- | --- | --- |
| Tool and vendor | name, personal or company licence, data contract | knowing who you have a contract with, and who you do not |
| Use | what it does, in which process | the same tool changes risk from one use to another |
| Owner | who introduced it and who answers for it today | putting a name on every use |
| Data | which data goes in, and of which class | deciding whether the use is allowed |
| People affected | customers, employees, candidates, suppliers | knowing whether notices or AI Act obligations apply |
| Check | who checks the result before it takes effect | knowing where an error can leave the company |
| Decision | continue, correct or stop, with a date | closing every row with a choice |

The NIST AI Risk Management Framework, the voluntary US reference, lists an inventory of systems among its governance practices (Govern 1.6) and asks you to plan how a system is decommissioned too (Govern 1.7), as the [framework core](https://airc.nist.gov/airmf-resources/airmf/5-sec-core/) puts it. For the list to be true, people using an unauthorised tool must be able to say so without consequences: Perplexity suggests a period of two to four weeks to declare it. **Every row of the list ends with a decision and a date; otherwise it is a census and governs nothing.**

## Who is in charge, and with what mandate?

One person who coordinates, with a written mandate from management and dedicated time, and a small group that decides on the cases that matter. The AI answers agree on this, and ChatGPT adds that the person can already be in the company: you do not need to hire a Chief AI Officer straight away. Who chairs changes: the chief operating officer for Google's AI Mode, the head of IT or technology for the AI Overview and for Perplexity. According to the Observatory, 54% of large companies are moving towards centralised governance.

| Role | Usually | What it decides |
| --- | --- | --- |
| Management | chief executive or general manager | the mandate, the acceptable risk, the cases that touch people or customers |
| AI coordinator | someone already in the company, with dedicated time | the list, the rules, the fast lane, the reviews |
| IT and security | head of information systems | allowed tools, access, integrations, activity logs |
| Legal and data protection | in-house counsel and data protection officer (DPO) | allowed data, vendor contracts, impact assessments |
| HR | head of people | notices to workers, training, relations with worker representatives |
| Process owner | whoever runs the department using AI | whether the use works, who checks, what happens when it gets things wrong |

NIST asks for roles, responsibilities and lines of communication to be documented and clear (Govern 2.1), and for executive leadership to take responsibility for decisions about AI risks (Govern 2.3). The group, though, should not approve everything. **A committee that reviews every request once a month becomes the reason people go back to personal accounts.** It decides on yellow and red cases; the rest flows.

## Who approves a new use of AI?

It depends on the risk of the use, not on the tool. The AI answers all propose three traffic-light levels and a fast lane: 5 working days for Google's AI Overview in the English question and for AI Mode in the compliance question. ChatGPT points out that the traffic light is an “internal classification”, not the AI Act's; true, but the two must talk to each other: a use that is high risk under the AI Act cannot be green for you.

| Level | When | Examples | Who approves |
| --- | --- | --- | --- |
| Green | tool already approved, public or internal data, a person checks before use | summaries, drafts, internal translations | nobody: you just add it to the list |
| Yellow | confidential or personal data, or the result reaches customers or suppliers | assistant over company documents, draft replies to customers | the coordinator, with IT and data protection |
| Red | decisions about people, uses in annex III of the AI Act, actions the AI takes without a person first | CV screening, employee evaluation, a chatbot answering customers on its own | the group, with legal and management |

The table is an example to adapt: your management sets the thresholds, on the cost an error would have in your process. ChatGPT, in the English question, adds a rule that prevents everlasting prototypes: “A prototype can be approved for a controlled test without being approved for production”. **Every trial starts with a threshold to reach and an end date, both set beforehand.** What the AI Act requires for red uses is in the guide on [what falls to those who use AI](/guides/ai-act-deployer-obligations).

**Would you rather talk it through?**

Book a call with people who work on these topics every day: we will look at your case together.

- [Book a call](https://vms.srl/booking)

## Which data can go into an AI tool?

It depends on the class of the data and on the contract with whoever provides the tool. If you have an information security policy, you already have the classes; the AI rule needs to add, for every approved tool, which classes it accepts.

| Class | Examples | Where it can go |
| --- | --- | --- |
| Public | website, catalogues, press releases | into any approved tool |
| Internal | procedures, minutes without personal data | into tools with a company licence that contractually excludes training on your data |
| Confidential | contracts, quotes, prices, projects | into company tools with role-based access |
| Personal | data on customers, employees, candidates | into tools with a data processing agreement, for a stated purpose |
| Special and secret | health, judicial data, trade secrets | only inside your perimeter, on private or on-premise infrastructure |

A company licence does not settle it. Google's AI Mode writes that Enterprise licences wipe out “the risk of data leak”; ChatGPT, to the same question, that paying for a company licence “does not automatically make any use safe”. The second is right. The contract excludes training on your data, but it does not decide who inside the company can see what: an assistant connected to company documents answers with whatever its permissions let it read, and if the permissions are broad, so are the answers. In the [searchable document archive](/case-studies/enterprise-document-rag) we built for an industrial company, 225 GB of regulations, HSE, safety and operations documents, the agent looks for the answer only in the documents the asker's role lets them see. Where to run the models for the most sensitive data is in the guide on [AI in the cloud or on your servers](/guides/ai-on-premise-vs-cloud).

Banning is not enough. The Observatory says so too: large companies know that banning generative AI “would lead to shadow AI and loss of control”, and only 24% ban tools not provided by the organisation. **If the company tool is worse than the personal one, the rule gets bypassed: the approved tool comes before the ban.**

Before checking who uses what, though, bring in HR and legal. Google's AI Overview, asked the compliance question, advises analysing network logs to find unauthorised AI, without naming the Italian rules on monitoring workers. [Article 4 of the Workers' Statute](https://www.normattiva.it/uri-res/N2Ls?urn:nir:stato:legge:1970-05-20;300~art4) lets you use the information collected only if the worker has received adequate information on how the tools are used and how checks are carried out, in line with privacy law; depending on the tool, a union agreement may also be needed. Monitoring is designed first and written into the policy.

## How do you measure whether AI works?

Against the starting point, with a number chosen beforehand. Among large companies using generative AI, only 54% try to measure its benefit, almost always by collecting employees' opinions; only 11% monitor it periodically and in a structured way (Observatory, February 2026). Google's AI Overview, in the English question, puts it in one line: “Never measure AI by "usage."” How many people open Copilot tells you how many open it, not what changed in the work.

| Use | What to measure | What is not enough |
| --- | --- | --- |
| Writing assistant | time per case and corrections before sending, against the same work without AI | how many people use it |
| Customer chatbot | requests closed without an operator, together with satisfaction and complaints | how many conversations it handles |
| Document assistant | correct answers on a sample checked by an expert, questions left unanswered | how many questions it receives |
| Agent preparing cases | corrections by the approver, closing times, cost per case | how many cases it prepares |

Next to the benefit you measure the cost: per case, not per licence, with spending limits per model and per period. An agent that works well but costs more than the work it replaces is not a result. **The number is chosen before you start, and on the set date it decides whether the use continues, changes or stops.**

## Who is accountable when AI gets it wrong?

Towards customers and third parties, the company; inside the company, the owner of the process the AI works in; whoever uses the tool answers for having checked the result before using it. The AI answers disagree. For Google's AI Mode “an AI error is a company error”; for the AI Overview “the end user remains solely responsible for the output”; in the English question, “The Business Unit Head who implemented the tool owns the operational risk”. Each has a piece of the truth; none separates the levels.

| Level | Who is accountable | For what |
| --- | --- | --- |
| Towards customers and third parties | the company | what the tool says or does in its name |
| Inside the company | the process owner | how AI is used: who checks, when to stop, what to do after an error |
| In daily work | whoever uses the tool | having checked the result, with the rules and training received |
| Towards the company | the tool's vendor | what the contract promises: availability, security, data processing |

The most quoted case is Canadian. In February 2024, Air Canada's chatbot had assured a customer that a bereavement fare could be obtained retroactively; the customer bought the ticket, and the airline then refused it. The British Columbia tribunal found a negligent misrepresentation and did not let the company disown what the chatbot had written (Moffatt v. Air Canada, 2024 BCCRT 149, as reported by [Agenda Digitale](https://www.agendadigitale.eu/industria-5-0/le-parole-dellia-possono-vincolare-limpresa-la-risposta-del-diritto/), in Italian). The same article warns that not every chatbot answer becomes a contractual commitment: content and the other information the customer had both count. But **telling a customer that the chatbot got it wrong is not a defence.**

The internal level needs people who can say no. For high-risk systems the AI Act wants human oversight assigned to people “who have the necessary competence, training and authority” ([article 26](https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-26), paragraph 2). For employment decisions Italian law is already further ahead: since 23 October 2026, final decisions on hiring, changing and ending employment, including disciplinary measures and performance evaluations, rest with a person who has real and autonomous power ([Legislative Decree 179 of 2026, article 41](https://www.normattiva.it/uri-res/N2Ls?urn:nir:stato:decreto.legislativo:2026-10-07;179~art41), in Italian).

Errors are then collected: a simple way to report them, an incident log, a person who can stop the tool. NIST asks for all three: user feedback to report problems (Measure 3.3), post-deployment monitoring with the ability to override and decommission (Manage 4.1), incidents and errors communicated, tracked and documented (Manage 4.3).

## How do you build the rules into the systems?

With controls that do not depend on people's memory. A rule written in the policy holds as long as someone remembers it; the same rule in the system always holds.

| Rule | Control in the system |
| --- | --- |
| everyone sees only the documents meant for them | the AI reads with the permissions of whoever asks |
| a person approves the actions that matter | the AI prepares, a person confirms before the change takes effect |
| an error must be traceable | a trace of the context, the data used, the steps and the cost of every operation |
| spending stays under control | limits per model and per period |
| data does not train other people's models | services that exclude it by contract, or private infrastructure |

This is how we work when we bring AI into established companies: before release come human approval on risky operations, the decision log and spending limits per model. In a group with several companies the separation holds between them too: on the [AI agent audit platform](/case-studies/agentic-ai-auditing-knowledge-platform) every organisation has its own isolated space, with separate data, definitions and contacts. **Automation goes as far as the cost of an error stays acceptable, and from there a person decides** ([how we ensure reliability](/frequently-asked-questions#affidabilita-ai)). Observability tools, which we always recommend and which are switched on at your decision, leave for every operation the trace you need to [reconstruct an AI decision](/frequently-asked-questions#tracciabilita-ai) even months later, and [your data does not train models](/frequently-asked-questions#dati-aziendali) belonging to others. For management, legal and compliance we run [training paths](/services/ai-enablement) on decisions and governance, built on your company's use cases.

## What does the law require, and what not?

Less than many answers say, and different things. To those who use AI, the AI Act asks for training measures for everyone, transparency in some cases and heavy obligations only for high-risk uses; Italian law adds notice to workers and a person's decision on employment. It does not require a committee, an AI officer or an internal list of systems.

| Rule | What it asks of those who use AI | Since |
| --- | --- | --- |
| AI Act, article 4 | measures supporting staff's AI literacy, without imposing a specific level | 2 February 2025, rewritten by Regulation (EU) 2026/1744 |
| AI Act, article 50 | disclosing deepfakes and generated texts of public interest; the chatbot presenting itself as AI falls to whoever provides it | 2 August 2026 |
| AI Act, article 26 | for high-risk uses in annex III: human oversight, logs kept for at least six months, informing workers before use at work | 2 December 2027 |
| Law 132 of 2025, article 11 | AI at work that is safe, reliable and transparent, without discrimination; notice to the worker in the cases of article 1-bis of Legislative Decree 152 of 1997 | 10 October 2025 |
| Legislative Decree 179 of 2026, article 41 | a person's final decision on hiring, changing and ending employment; notice to the worker before using the system | 23 October 2026 |
| Legislative Decree 179 of 2026, article 42 | AI affecting organisation, pace or the way of working goes into the risk assessment; information and training for workers on its risks | 23 October 2026 |

[Article 4](https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-4) is in force since 2 February 2025 and rewritten by Regulation (EU) 2026/1744: measures supporting staff's AI literacy, without imposing a specific level. Almost every AI answer still describes it as a duty to train staff, and none says it was rewritten; only Perplexity lists the regulation that changed it among its sources. According to the Observatory, more than one large company in two has started AI literacy initiatives, but only 15% have a structured AI Act compliance project, integrated with the other rules. [Article 11 of Law 132 of 2025](https://www.normattiva.it/uri-res/N2Ls?urn:nir:stato:legge:2025-09-23;132~art11) and [article 1-bis of Legislative Decree 152 of 1997](https://www.normattiva.it/uri-res/N2Ls?urn:nir:stato:decreto.legislativo:1997-05-26;152~art1bis), both in Italian, require informing workers when an automated system decides or monitors something about their work.

Three claims found in the AI answers need correcting. Google's AI Mode writes that the AI Act requires creating an internal register: for a private company that uses AI it does not; article 26 asks you to keep the logs of high-risk systems, which is something else. The same surface presents the fine of up to 15 million euro or 3% of turnover as the price of operating without rules: in [article 99](https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-99) it is the ceiling for breaching the deployer obligations of article 26 and the transparency obligations, not for lacking a policy. And in the English question it allows “the free tier of ChatGPT” in the green level, which has no company contract on data.

**The list, the group and the policy are not obligations: they are the tools with which you prove, on the day it matters, that you met the real ones.** The detail of the obligations, case by case and with a self-check, is in the guide on the [AI Act for deployers](/guides/ai-act-deployer-obligations).

## Do you need ISO/IEC 42001?

Not to get started, and it is not mandatory. ISO/IEC 42001, published in December 2023, “specifies the requirements and provides guidance for establishing, implementing, maintaining and continually improving an AI management system”, for organisations that provide or use products or services built on AI ([standard page on AFNOR](https://www.boutique.afnor.org/en-gb/standard/iso-iec-420012023/information-technology-artificial-intelligence-management-system/xs142381/355348)). The [NIST AI Risk Management Framework](https://www.nist.gov/itl/ai-risk-management-framework), published on 26 January 2023, is voluntary and free, and organises the work into four functions: govern, map, measure, manage.

| Question | ISO/IEC 42001 | NIST AI RMF |
| --- | --- | --- |
| What it is | a standard with the requirements of an AI management system | a voluntary framework for managing AI risks |
| Who publishes it | ISO and IEC | NIST, the US standards institute |
| Can you certify | yes, with a certification body | no |
| What it costs to read | you buy it | free |
| What it is for | proving to a third party how you govern AI | a checklist so nothing gets forgotten |

Among the AI answers it is mentioned by ChatGPT and Perplexity, which proposes 42001 as the structure and NIST as the checklist; none says when certification pays off. Our reading: it pays off for those selling AI systems to customers who ask for it, and for those who already have a certified management system, such as ISO 27001, because the work adds to something that exists. **For everyone else, the list of uses, written roles and an incident log are worth more than a certificate.**

## Where do you begin?

With the list, not the policy.

1. Make the list of uses across all departments, with a few weeks in which unauthorised tools can be declared without consequences.
2. Appoint the coordinator, with a written mandate from management, and the group: IT, legal, data protection, HR, the process owners.
3. Write the data classes and, for every approved tool, which ones it accepts; give the company tool to whoever needs it before banning the personal one.
4. Set the three approval levels, with a fast lane for green and an end date for every trial.
5. For every yellow or red use, choose the number to measure, the starting point and the decision date.
6. Open the incident log and a simple way to report an error.
7. Write the policy, on what you have decided, and build the training on it.

**The policy is written last, because it describes decisions already taken; written first, it is a list of good intentions.** If instead AI is not in your company yet and you are looking for the first process to start from, read the guide on [AI in business: where to start](/guides/start-with-ai-in-business).

## An example: a company with 600 people

A company with 600 people has ChatGPT in several departments, Copilot in administration, a customer service chatbot and two prototypes built by IT. It is the case we put to Google and the AI assistants; the details that follow are invented, to be replaced with yours.

The first month produces the list: about forty uses, half on personal accounts. Three upload customer contracts into a tool without a data processing agreement: they stop at once, and the people using them get the company tool the same week. Management appoints the operations manager as coordinator, with a written mandate and one day a week of dedicated time; the group meets every two weeks, only on yellow and red cases.

The customer service chatbot is red: it answers on its own, and what it promises binds the company. It stays on, but answers only from approved documents, hands refunds and complaints to an operator, and every week someone checks a sample of conversations. Of the two prototypes, one has a threshold and a date and becomes a project; the other, with no number to improve, stops.

HR asks to use AI for a first screening of CVs. It is a use in annex III of the AI Act, with the obligations of article 26 from 2 December 2027, and the group treats it as red. Article 41 of Legislative Decree 179 of 2026 does not count searching for and screening candidates among final decisions; when AI enters the decision to hire, performance evaluations or disciplinary measures, though, the final decision already rests with a person today. After three months the company knows how many uses it has, who answers for them, which data goes where and whether the most important uses work. Only then does it write the policy: a few pages, on decisions already taken.

## Frequently asked questions

### Do you need a Chief AI Officer?

Not to get started. You need one person with a written mandate from management, dedicated time and a group with IT, legal, data protection and HR; often that person is already in the company. A dedicated role makes sense when AI uses become many and touch the product or customers.

### Should you ban ChatGPT at work?

Ban personal accounts, yes, if company data goes into them; ban AI, no. According to the Observatory of Politecnico di Milano, large companies know a ban leads to hidden use, and only 24% ban tools not provided by the company. First you provide a company tool with a data contract, then you ban the personal one.

### Does the AI Act require a register of AI systems?

Not from a private company that uses AI. For high-risk systems article 26 asks you to keep the logs the system generates for at least six months, which is something else. The list of uses is still the most useful tool for governing AI and for proving, on the day it matters, what you use and under which rules.

### Who pays if the chatbot gets it wrong with a customer?

Towards the customer, the company: in Moffatt v. Air Canada, in 2024, a Canadian tribunal did not let the airline disown what its chatbot had promised. Inside the company the process owner answers for it; towards the company, the vendor, within the limits of the contract.

### Do you have to tell employees about the use of AI?

Yes, when an automated system decides or monitors something about their work, such as hiring, tasks, evaluations or surveillance: article 11 of Law 132 of 2025 and article 1-bis of Legislative Decree 152 of 1997 require it, and since 23 October 2026 article 41 of Legislative Decree 179 of 2026 too. For AI that changes the way people work, article 42 of the same decree also requires information and training on its risks.

### Is ISO/IEC 42001 certification worth it?

It is worth it if you sell AI systems to customers who ask for it, or if you already have a certified management system such as ISO 27001. To start governing AI you do not need it; the list of uses, written roles, data rules and an incident log come first, and the standard, if it comes, builds on them.

## AI in every department and no common rules?

Tell us what you use today, in an introductory call, free and with no commitment: we tell you where to start and which rules are worth building into the systems.

- [Let's talk about your project](/en/contacts)

## Articles on the same topics.

*Keep reading*

- [Data quality: data decides, not the model](/en/insights/data-quality-why-ai-fails-on-dirty-data.md): When AI gets it wrong, blame the data, not the model. How to identify, sanitize and refresh your sources so AI actually works in your next business project.
- [European AI Act who does it really affect?](/en/insights/eu-ai-act-deployer-business.md): The EU AI Act doesn't only target those who build the models, but those who use AI in business. What applies from 2 August 2026 and what moves to 2027-2028.
- [Small is the new big: SLM for enterprise](/en/insights/small-language-models-enterprise.md): Enterprise AI doesn't need giant models. Learn how Small Language Models (SLMs) cut inference costs, reduce latency, and ensure complete data sovereignty.