Guide · Last updated 19 October 2026
Where to start with AI in a mid-size company
Start with one concrete problem that costs you time every week, pick a single project as the pilot and decide in advance how you will know it works. Do not buy more tools first. A small pilot with a stop point after each step gives you answers within a few weeks and lowers the risk that the effort fizzles out.
Start with the problem, not the tool
Most mid-size companies already have several systems. What is usually missing is that the information connects. You will recognise it if you call two people to find out how a project is going, if the same task is done by hand every Friday, or if important decisions sit in a chat group only a few people read.
Write down three situations like that. The one that costs the most time, or causes the most worry, is your candidate.
Pick a pilot project
A good first project meets five conditions:
- It happens every week, so you see the effect quickly.
- It has an owner, someone in the company who cares about the result.
- It can be measured, for example time, number of reminders or number of calls.
- The risk is low, no sensitive decisions are automated.
- Someone wants it, start where there is will, not where the biggest gain is on paper.
What a phase looks like
| Step | What happens | Result |
|---|---|---|
| 1. Discovery | A few conversations about how work really runs. | A short plan and a fixed price for the next step. |
| 2. Pilot | One real project and an agent that helps, from the first day. | Something that gets used. |
| 3. Stop point | You review the result together. | Continue, change course or stop. |
| 4. Widen | More projects and more features. | At the pace people actually use them. |
What it costs and how to limit the risk
Ask for a fixed price per step, with a cap, and a clear process for changes. Then you know what each step costs before it starts, and you can stop at every stop point. An hourly rate fits new requests best, not the core work.
What decides whether it gets used
The technology is rarely the hard part. What decides it is whether your staff open it. It helps when it is short and in the right language, when it needs no new login and when someone in the company owns it. Measure use, time saved and how confident your staff feel, not the number of features.
Data protection
Sign a data processing agreement before anything starts. Decide which data the agent may see, and start with information that is not sensitive.
Common questions
Do we have to switch systems?
No. A good solution is built on top of the systems you already have.
How long does a pilot take?
Think weeks, not years. The length is set in discovery.
Who should own the pilot?
Someone in the company who cares about the result, not only a supplier.
What if our staff do not want to use it?
Start with those who do and build around their needs. What they find awkward becomes input for the next step.
Do we need our own technical staff?
No, but someone on your side needs to be able to say what works and what does not.