IAM: Access management in SMEs, from employees to AI agents
In short: IAM (Identity and Access Management) is about knowing who each user is, what they're allowed to access, and when that access should end. MFA is only the login step, and SSO (single sign-on, one login for multiple systems) is a convenience, whilst IAM is the full record of who has access to what. The Danish Data Protection Agency describes the principle this way: an employee should have access only to the data and processing that their job requires, and that access should close when the need ends. The same principle should apply to an AI agent.
Table of contents
If you've ended up here, it's perhaps because a reviewer, a customer or your insurer has asked who has access to what in your systems, and the answer was "we're not entirely sure". It's common in companies that have grown quickly: some people have been given access because it was easy, and no one has tidied up when people changed roles or left.
What is IAM, and what does it do?
IAM covers the whole life of an access. First, a person or a system needs an identity. Then that identity is given rights to specific systems and data. Finally, those rights need to change when tasks change, and be removed when the person leaves. The Danish Data Protection Agency uses the term "rights management" and explains it with a boarding pass: you get access to one flight, one departure, to the cabin and not the cockpit, and the card no longer works after the journey ends.
Typical functions:
- a single overview of users and their access
- assigning rights by role, so a new salesperson gets the same as the other salespeople
- a person responsible for each system who approves who gets access
- rapid closure of access when someone leaves, goes on leave or changes tasks
- special oversight of administrators and others with broad rights
- temporary accounts for external people, temporary workers and consultants
- logging and regular reviews of who has what
IAM, SSO and MFA are not the same
| Term | The question it answers | Example |
|---|---|---|
| IAM | Who are you, what are you allowed to do, and when should it stop? | A list and a process for who has access to accounts, customer data and files |
| SSO | Can I use just one login for multiple systems? | You log in once in the morning and are then inside your email and business applications |
| MFA | Is it really you logging in? | Password plus confirmation on your phone |
The three complement each other. SSO without good rights management gives a convenient path into everything. MFA without rights management ensures that the right person comes in, but not that the person can only see what they need to.
A well-known example from the public sector is MitID Erhverv: the organisation itself creates its employees as users, gives them rights to specific self-service functions and can give an accountant or administrative bureau a business mandate. There are five roles, and most employees are simply ordinary users, whilst a few have responsibility for administering the others.
When do you need IAM?
The signs are clear: you share logins, no one can say off the top of their head who has access to the accounts, former employees still have access, or you use many systems, each with its own user list. The Danish Data Protection Agency's guidance mentions failure to close access when someone leaves and misuse of administrator rights as two of the breaches it reviews.
If you're five people with three systems, a shared list of access and a checklist for new and departing employees often suffices. The need for a proper IAM system grows with the number of systems, people and external users.
For organisations covered by the NIS 2 Act, access control policies and asset management are among the measures that must be in place as a minimum, and the use of multi-factor authentication is mentioned alongside them.
When people change roles or leave
The Danish Data Protection Agency points out that leave, parental leave, sick leave and suspension require the same attention as someone leaving: you must decide which access should be withdrawn or temporarily deactivated, and rapid closure normally requires that you're prepared. For external people such as consultants and temporary workers, the guidance mentions measures such as temporary user accounts, separation of duties and minimisation of privileged access rights. The guidance also notes that it matters for speed whether rights management is decentralised, where multiple people create access themselves, or centralised, so there's one place to close things down.
What does it cost?
- licence per user for an identity solution, if you choose one
- time to map out your current access and tidy up, which is often the largest item
- integration with systems that don't support common login
- connection to your HR system, so access follows employment, see for example HR systems
- training for system administrators in approving and reviewing access
- ongoing management: reviewing rights and handling exceptions
Buy a standard system or build your own?
Standard, almost always. The identity part is security-critical and shouldn't be built yourself. First check whether the systems you already use have built-in role and user management, as that can go a long way for an SME.
A customised layer or an integration layer makes sense when you have business applications without good user management, where access is currently created and closed by hand. Here, an automatic link between your HR system and your other systems can remove the risk that someone forgets to close an account.
Whatever you choose, you should ask: can we see who has access to what on one overview? Can we close an employee's access to all systems in one go? And can we see who approved a given access?
Where does an AI layer fit in?
AI agents are new users in your systems, and they should be treated as a new employee with very limited rights. The Danish Data Protection Agency's catalogue mentions measures such as data access on a need-to-know basis, minimisation of privileged access rights, separation of duties and temporary user accounts. These can all be transferred directly to AI agents:
- a separate identity for each agent, never an employee's login
- read access first, write access only where the task requires it
- no administrator rights
- access limited to the data the task uses, and time-limited where it makes sense
- a named person responsible for each agent and an easy way to switch it off
On top of IAM itself, AI can prepare the half-yearly review by gathering lists of access and flagging ones that look unusual, or filling in creation forms based on a new hire. AI can't decide who should have access — that's a manager's or system owner's job — and it shouldn't be able to give itself more rights. Read more about AI and data security.
How to approach it
- Make a list of your systems and who has access to each of them. It's the job that takes the longest, and it gives the most.
- Appoint someone responsible for each system, ideally the nearest manager or system owner, as the Danish Data Protection Agency mentions.
- Define a few roles, and give each role only the rights it needs.
- Write a short checklist for new employees, departing staff and those on leave, and link it to your HR system.
- Give administrators separate accounts, and turn on multi-factor login for them first.
- Set up a regular rhythm for reviewing access, and include AI agents on the list.
Agent360 can help you choose a solution or build the integration layer that links your HR system to your other systems: get in touch.
Sources
- The Danish Data Protection Agency, Rights Management and the guidance Styr på rettighedsstyring (December 2023): https://www.datatilsynet.dk/regler-og-vejledning/behandlingssikkerhed/rettighedsstyring (accessed 9 October 2026)
- MitID Erhverv, What is MitID Erhverv: https://www.mitid-erhverv.dk/om-mitid-erhverv/hvad-er-mitid-erhverv/ (accessed 9 October 2026)
- Act on measures to ensure a high level of cybersecurity (NIS 2 Act), Act no. 434 of 6 May 2025, section 6: https://www.retsinformation.dk/eli/lta/2025/434 (accessed 9 October 2026)
Frequently asked questions
What's the difference between IAM, SSO and MFA?
IAM is the full record of who has access to what, and when that access should end. SSO means one login for multiple systems, and MFA confirms that it's really you logging in, for example with a password plus confirmation on your phone. The three complement each other.
Does a small company need an IAM system?
Not necessarily. If you're five people with three systems, a shared list of access and a checklist for new and departing staff often suffice, and the need for a proper IAM system grows with the number of systems, people and external users.
What rights should an AI agent have in our systems?
As few as possible. Give each agent its own identity and never an employee's login, start with read access, give write access only where the task requires it, and give no administrator rights. Each agent should have a named person responsible for it and an easy way to switch it off.
Contents
What takes up your time?
Write down the task that takes up the most time. We reply whether it is worth building, and we say so if it is not.
Write what takes up your timeRelated articles

Billing address: what it is, and what your invoice must include
Understand billing addresses, invoice requirements, and when to use separate billing and delivery addresses for your business.

Simple accounting: what's legal for a small business?
Learn when simple accounting is legal for small businesses in Denmark. Understand the requirements for digital bookkeeping, invoicing, and document retention.

Annual financial statements: how to prepare and file them
Learn what an annual financial statement includes, who prepares it, and how to file it with the authorities.