AI governance for SaaS companies: one register for the AI in your product and the AI in your company.

I'm Yiannis Papastergiou, an independent AI consultant based in Thessaloniki, Greece. For SaaS companies of roughly €1M+ revenue I list every AI system you ship and every one your team uses, place each in its EU AI Act risk tier, and write the policies, controls and records a customer's security team can read. In person in Athens and Thessaloniki, remote everywhere else.

AI governance · SaaS companies · EU AI Act · 3–6 weeks

About 3–5 minutes, mostly one-tap answers, then your name and work email. Your result on screen, with a 7-day plan. No call required.

Take the free AI Growth GPS

See the AI P.R.O. Method

What is AI governance for SaaS companies?

AI governance for SaaS companies is the set of records and rules that shows what AI is in your product, what AI your team uses, which EU AI Act role and risk tier each system carries, and who owns each control. A SaaS company needs it twice: as the provider of its own AI features, and as a deployer of other people's tools.

With me the work takes three to six weeks. Your investment is confirmed in a written proposal. It is not legal advice or a certification.

See the AI P.R.O. Method

The short answer

Why does AI compliance for SaaS companies reach the sales pipeline before the regulator?

A regulator may never write to you. A customer's security team will. The AI section of a vendor questionnaire asks which model providers receive customer data, whether that data trains any model, how outputs are checked and logged, and where the product sits under the EU AI Act (Aetos Data Consulting, 2026).

The frameworks behind those questions are published. The Cloud Security Alliance's AI Controls Matrix (July 2025) has 243 controls across 18 domains and a companion self-assessment questionnaire.

The Act's heaviest duties start in December 2027. Your next enterprise deal starts with a questionnaire.

Inside a SaaS company it tends to look like this:

The AI feature shipped in a sprint. Nobody wrote down which model it calls, what it sends, or what happens when the model is wrong.

The sub-processor list names the cloud host and the email provider, and not the model API.

Engineers use coding assistants, support has a drafting tool and sales has a note-taker on every call. No one holds the whole list.

The problem

Where does AI governance for SaaS have to reach?

Five places AI sits in a software company. Each gets a register row, a risk tier and an owner.

You built it and it carries your name, so under the EU AI Act you are its provider. The register records the model behind each feature, the data it receives, what users are told and how wrong answers are caught.

AI features in the product

Most SaaS features call a general-purpose model from another company. Since 2 August 2025 that model's provider has owed you documentation about it (Article 53). The register notes what you hold, which features send customer content to the model, and whether any of it trains anything.

The model API and your customers' data

Coding assistants, drafting tools and note-takers: which repositories and records they may read, and whether secrets or customer data can reach them.

Engineering and internal tools

An assistant that talks to your customers has to be designed so people know they are dealing with AI (Article 50).

Support and success

A tool that screens or ranks job applicants is a high-risk use under Annex III. If you use one, it is the internal entry that needs the most careful record.

Hiring and people

The register names the control each system requires. The records your team keeps afterwards show the control is working.

Five places

EU AI Act compliance for SaaS companies: what applies, and when?

Article 3 calls you a provider when you develop an AI system and place it on the market under your own name, and a deployer when you use one under your authority. A SaaS company is both: provider of the features it ships, deployer of the tools its staff use.

You hold two roles at once

Banned practices have applied since 2 February 2025 and transparency rules since 2 August 2026. Regulation (EU) 2026/1744, in force since 27 July 2026, moved high-risk obligations for Annex III uses such as hiring and credit to 2 December 2027. Marking of AI-generated content has a grace period to 2 December 2026 for systems already on the market (European Commission and Cooley, August 2026).

The dates, after the 2026 Omnibus

Annex III lists uses, not technologies: recruitment, evaluating students, credit scoring. Under Article 25, whoever changes a system's intended purpose so that it becomes high-risk is treated as its provider. Your documentation and terms should say what each feature is for, and what it is not for.

A customer's use can change your tier

This is information, not legal advice. Your lawyer confirms the edge cases.

What applies now

Who should build AI governance for a SaaS company?

Option

Best at

Check first

This table is my own judgement. A common pairing is someone to build the register and framework, and a lawyer to check the legal conclusions.

Your security or GRC lead

Knows the systems and already runs SOC 2 or ISO 27001 evidence.

Whether they have the weeks, and who classifies under the EU AI Act.

Compliance automation platform

Collecting evidence and tracking controls once you know what they are.

Software records decisions. Someone still decides each feature's role and tier.

Law firm

Legal interpretation, customer contracts and speaking to a regulator.

Who builds the register and the rules engineers will follow.

Independent AI consultant (me)

Finding what is shipped and used, classifying it and writing a framework a product team will follow.

No legal opinion and no certificate. One person.

Your options

Who would be writing your SaaS company's AI governance?

One named person: Yiannis Papastergiou, independent AI consultant, based in Thessaloniki, Greece.

As Product Owner at Natech I led its anti-money-laundering software from desktop to a modern web platform. My product background is regulated fintech, where "trust us" isn't an answer.

Product owner in regulated fintech

At Companies House, the UK's official business registry, I worked on the systems behind millions of searches and filings every year, and was security and quality champion.

Security and quality champion

At ARAG UK, a 500-person legal insurance business, I started as the sole developer and became technical lead.

Every seat in a software team

I'm not a lawyer and I don't issue certificates. My services are new as a set, and I have no SaaS governance results to publish.

Who you'd be dealing with

AI governance services for SaaS companies, and the investment

About 3–5 minutes, mostly one-tap answers, then your name and work email. Your result on screen, with a 7-day plan. No call required.

Take the free AI Growth GPS

Your investment is confirmed in a written proposal.

See the AI P.R.O. Method

Your investment is confirmed in a written proposal.

System inventory across product and company, risk classification, a governance framework and a board briefing. 3–6 weeks.

AI Governance & EU AI Act Compliance

The governance work

The staff side of the usage policy: hands-on sessions on real work. Your investment is confirmed in a written proposal.

AI Training & Executive Workshops

One owner for AI as the product changes. Your investment is confirmed in a written proposal.

Fractional Chief AI Officer

Often taken with it

Not yet if your product has no AI in it and your team uses none. The inventory would be empty.

Services and investment

How does the work run with a SaaS team?

Shown at six weeks. The register alone: three weeks. Stop after any step and keep what you hold.

your growth constraint, your AI readiness and a 7-day plan, on screen

Step 1

The free AI Growth GPS: a 3–5 minute conversation about your SaaS company.

the investment, confirmed in writing

Step 2

A written proposal, with the scope and the investment confirmed before anything starts.

every AI system listed, with its role

Weeks 1–2

System inventory. You point me to every AI system in the product and the company.

each system in its EU AI Act risk tier

Week 3

Risk classification.

policies, controls, documentation

Weeks 4–5

The framework.

Week 6

Board briefing.

No call required

Questions SaaS founders ask about AI governance

Almost always, in part. If your product has an AI feature, you are its provider. If your staff use AI tools, you are a deployer. A feature that talks to people must make clear it is AI (Article 50, in effect since 2 August 2026). High-risk duties apply only when a feature is intended for an Annex III use such as hiring or credit scoring, from 2 December 2027.

Does the EU AI Act apply to a SaaS company?

It is a good base, and it doesn't answer the AI questions. Security frameworks show how you protect systems and data. They don't record which model each feature calls, whether customer data trains anything, what users are told, or which EU AI Act tier a feature sits in. I write the register to sit beside the evidence you already keep.

We already have SOC 2 or ISO 27001. Isn't that AI governance?

Only if customers ask for the certificate. ISO/IEC 42001:2023 sets requirements for an AI management system, and a certificate comes from a certification body after an audit. I am not one. If a customer insists on the certificate, the register, the classification and the written policies are a starting point for that work, not a substitute.

Does a SaaS company need ISO/IEC 42001 certification?

Then the tier may change, and so may who carries the provider duties. Annex III lists recruitment and credit scoring as high-risk uses, and Article 25 treats whoever changes a system's intended purpose into a high-risk one as its provider. State the intended purpose in your documentation and terms. I draft the register entry. The contract wording is work for your lawyer.

What if a customer uses our feature for hiring or credit decisions?

It can. Article 2 covers providers that place AI systems on the EU market wherever they are established. If you have EU customers, classify the features they use. I work with companies outside Greece remotely, and I have no office or team in the UK or the US.

We are a UK or US SaaS company. Does EU AI Act compliance apply to us?

Plan starts from €1,500 (the Numbers Round Table, refunded if it isn't worth it within seven days). Everything after that is quoted as one fixed figure in a written proposal, sized to your business, before anything starts. Scope sets where you land: the register and classification alone, the full framework, or an operating model with staff training.

What is the investment for AI governance for a SaaS company?

Three weeks for the register and classification, six for the full framework and board briefing. From you: time with the people who own each AI feature and each internal tool, a plain account of what each is used for, and a place on your board's agenda. Interviews run by video call, in English or Greek.

How long does it take, and what do you need from our team?

A named owner on your team, usually whoever already runs your security or compliance evidence. The framework includes one standing rule: a new AI feature or tool gets a register row before it ships. If you would rather have that owned from outside, the Fractional Chief AI Officer service is monthly, with no contract.

Who keeps the register current once the product changes?

Questions

Not sure governance is your first move? Start with a free 3–5 minute conversation about your SaaS company.

About 3–5 minutes, mostly one-tap answers, then your name and work email. Your result on screen, with a 7-day plan. No call required.

Take the free AI Growth GPS

A guided conversation with my AI Growth GPS. Tap or type your answers. Why it's free: you're pitched by AI consultants every week, so you get something useful before we ever speak. Nothing is shared.

Discover · The free AI Growth GPS

When the questionnaire asks what your AI does, hand over a page instead of a guess.

Take the free AI Growth GPS, or see how the AI P.R.O. Method works first. Neither starts with a call.

About 3–5 minutes, mostly one-tap answers, then your name and work email. Your result on screen, with a 7-day plan. No call required.

Take the free AI Growth GPS

See the AI P.R.O. Method