Agentverkstad
Menu

Guide · Last updated 11 November 2026

AI strategy for a small team

An AI strategy is a short plan for what you will use AI for, who owns it and what data never leaves the house. Every team that wants to start with AI without trying everything at once needs one. For a small team it fits on one page and answers four things: which work should get easier, who owns it, which data never leaves the house and how you see that it works. Start with a single workflow, try it for a few weeks and widen only when it is being used.

The four questions

Pick one workflow

A good first workflow comes back every week, has a clear owner, can be measured and carries low risk. It should be something one person in the team wants help with, where the gain shows already in daily work.

A small team is often short of time, so it pays to pick a workflow that fits inside the normal week. That works better than starting a project beside the work.

Rules for data

Write three rules on the same page. Which data the agent may see. Which never leaves the house. Who approves before anything goes out. If the agent can reach personal data, a data processing agreement should be in place before you start.

What the page can look like

Common pitfalls

The most common is starting several workflows at once, so that none of them gets any time. Next comes having no owner, and forgetting the measurement. The third is that the rules for data are written only after something has gone wrong. Write them on the first page, and you avoid doing it afterwards.

A rhythm that holds

Set a fifteen-minute review every week for the first weeks. Look at the measure, discuss what was awkward and decide one thing to change for next week. After a couple of weeks you can see whether the workflow works or needs to change.

When you widen

Widen only when the first workflow is used without reminders. Then pick the next one, with the same four questions. The strategy grows that way, one page at a time, and you avoid a large programme that nobody has time to follow. Also write down what you learned from the first workflow, so that the next one starts from what you already know.

If the work needs to connect across your systems, we help with that in Agentic Operations, where we map how the work flows and build the agents that take over the manual parts. The price is fixed per step, with a stop point after each. If you are one person who wants to start with your own agent, there is Your first AI agent.

Common questions

What is an AI strategy?

A short plan for what you use AI for, who owns it and what data never leaves the house. For a small team one page is enough.

Does a small team need an AI strategy?

Yes, a short one. Without it you get many small attempts that nobody follows up. One page with four answers is enough.

How many workflows should we start with?

One. When it is used without reminders, pick the next.

Who in the team should own it?

The person who benefits most from the result and has time to follow it. Name a deputy as well.

How do we protect our data?

Decide what the agent may see, start with what is not sensitive and sign a data processing agreement if personal data is involved.

When should we get outside help?

When the work needs to connect across several of your systems, or when you want to move fast and have no time to build it yourselves.

How long does it take to write the strategy?

The first version can be written in an hour, once you have chosen the workflow. You learn the rest while the workflow runs.

Read next

Tell us where the time goes.

Book a first call straight away, 30 minutes and free of charge. Or leave a few lines and we will get back to you.

Fetching available times …

Other ways in