How to govern AI in your company: a 4-step model
Governing AI in a company means knowing which AI systems are in use, who is accountable for each one and which controls they run under. The model we apply at Vivid Vision does it in four steps: pre-assessment, inventory and risk framework, operation with the company's team, and handover.
By Rafael Tamames, founder of Vivid Vision
What does governing AI mean?
AI governance is the set of policies, owners, processes and controls an organisation uses to decide which artificial intelligence systems it uses, oversee them and answer for their results and risks. A framework that works rests on six pieces.
- AI policy
- Principles, acceptable uses and red lines, with leadership's commitment.
- System inventory
- A single register of every model in use, in-house and third-party, with its purpose, its data and its risk level.
- Governance committee
- Approves use cases, reviews risks and settles what no single department can decide alone.
- Risk management
- Identifies and addresses bias, security, privacy, compliance and impact on people.
- Controls and traceability
- Documentation, human oversight, logs and vendor management, so everything can be audited.
- Metrics
- Inventory coverage, risks addressed, incidents and adoption.
What are the four steps to govern AI?
This is the process of Operate , the phase of Vivid Vision's AI Operating Partner devoted to governance. The order matters: each step builds on the previous one.
-
Pre-assessment
We map your AI systems in production, their risks and your priorities under the AI Act and DORA. In weeks, not months.
The output is a risk map with actionable priorities. It is free and with no commitment, and it helps decide whether it makes sense to operate together.
-
Inventory and risk framework
We build the single inventory and the governance model: risk classification, owners, controls and traceability.
It is the basis for everything else. Without knowing which AI systems make decisions today, including third-party ones and those teams use on their own, you cannot classify risk or assign owners.
-
Operation with your team
We deploy governance operating inside your organisation, prioritising the highest-risk systems.
Deployment is incremental. Impact is measured in NPS, P&L and adoption, not in deliverables nobody implements.
-
Handover
The operating model and the know-how stay with you. We install capability, not dependency.
We design transferable operating models with an executor-agnostic methodology, so the company keeps governing its AI on its own.
More detail: Request your pre-assessment
Who should lead AI governance?
It needs clear executive responsibility. A medium-sized company can start by assigning it to an executive and a committee; a large regulated corporation will probably need a dedicated CAIO.
| Role | What it is responsible for |
|---|---|
| Managing director or CEO | Assigns executive responsibility and commits leadership to the AI policy |
| CAIO or a leadership team member (CIO, CDO or COO) | Holds executive responsibility for AI strategy, risk and governance |
| Governance committee | Approves use cases, reviews risks and settles what no single department can decide alone |
| Business owner of each system | Answers for the use of a specific system in production |
| Risk, compliance and legal | Act as the second line of control |
More detail: Guide: AI governance framework, roles and CAIO
Which regulations set the pace for AI governance?
- AI Act
- Requires human oversight and accountability for high-risk systems. Annex III obligations apply from 2 December 2027. Full timeline in The AI Act for companies: what applies now and what comes in 2027 and 2028 .
- DORA
- Digital operational resilience for EU financial entities and their critical technology providers, fully applicable since 17 January 2025. It covers the AI systems that support critical processes.
- ISO/IEC 42001
- The certifiable international standard for an artificial intelligence management system. It provides the management framework that makes AI Act compliance easier. More in the ISO 42001 guide.
How long does it take to get AI governance running?
It depends on the size of the company and its regulatory exposure. The pre-assessment maps the AI systems in production, their risks and the compliance priorities in a few weeks. From there the governance model is deployed incrementally, starting with the highest-risk systems under the AI Act and DORA.
Which mistakes should you avoid when governing AI?
- Governing on paper
- A report nobody implements is not an operating model. Governance is operated with the team, inside the company.
- Letting each department decide alone
- Without shared governance, each department tests tools on its own and pilots never reach the business.
- Starting with the tool
- The first step is the inventory, not buying a platform.
- Creating dependency
- If the company cannot carry on alone at the end, the handover is missing.
Frequently asked questions
What is the difference between AI governance and complying with the AI Act?
The AI Act is a legal obligation. AI governance is the operating model the company uses to comply with it and, beyond that, to decide which AI it uses and how it measures the results.
Do I need a CAIO to govern AI?
It depends on size and regulatory exposure. A large regulated corporation will probably need a dedicated CAIO; a medium-sized company can start by assigning the responsibility to an executive and a committee.
Does third-party AI need governing too?
Yes. The inventory includes third-party systems and those teams use on their own. Anyone who uses an AI system in their professional activity is responsible for deploying it.
What does the pre-assessment include?
A working session in which we review your AI systems in production, identify gaps against the AI Act, DORA and Spanish regulation, and give you a risk map with actionable priorities. Free and with no commitment.
Further reading
Sources
Informational summary, not legal advice.
Take the first step with a pre-assessment
We map your AI systems in production, their risks and your priorities under the AI Act and DORA. In weeks, free and with no commitment.
Request your pre-assessmentKeep reading
What is an AI Operating Partner (and how it differs from a consultancy)
What an AI Operating Partner does, how it works and how it differs from an AI consultancy.
Read the pageThe AI Act for companies: what applies now and what comes in 2027 and 2028
Which parts of the AI Act already apply, what comes in 2027 and 2028 and which decisions a CEO needs to take.
Read the page