Three Places AI Actually Pays Off in a 20-Person Company

Most of the AI advice floating around is written for companies with a hundred engineers and a department for everything. If you're running a 20-person business in Norway, that advice is mostly noise.

I've spent the last year helping smaller companies figure out where AI actually moves the needle for them. Three patterns keep showing up.

1. The inbox and the calendar

This is the unglamorous one, and it's the one that pays back fastest.

A founder or daglig leder in a 20-person company spends a wild amount of time on email, scheduling, follow-ups, quotes, and small written replies. Copilot in Outlook, or even just ChatGPT in a browser tab, knocks the time on that work down by half. Maybe more.

Nobody writes a case study about it. But if you save the leadership team five hours a week each, that's the cheapest hire you'll ever make.

2. Quotes, proposals, and the first draft of anything

Small companies lose deals because the proposal didn't go out fast enough, or because the one person who writes well was on holiday.

AI doesn't replace the person who knows the customer. It gives them a passable first draft in two minutes instead of two hours. The human still edits, still adds the judgment, still hits send. But the slow part is gone.

Same goes for tender responses, internal docs, customer onboarding emails, and the quarterly update nobody wants to write.

3. Knowing what's in your own files

Most small companies have a SharePoint, a OneDrive, or a shared Google Drive that nobody can find anything in. The information is there. Finding it is the problem.

Copilot connected to your Microsoft 365 tenant, or a small custom setup pointed at your shared folders, turns that mess into something searchable in plain Norwegian or English. Ask it what was agreed with a customer two years ago. Ask it which template you used for the last NDA. Get an answer in seconds.

This is the one that quietly changes how the company runs.

What's not worth it (yet)

For a 20-person company, skip these for now:

• Custom-built AI agents that need a developer to maintain
• Anything sold as a "transformation initiative"
• AI features bolted onto a tool you weren't going to use anyway

Start with the three above. Get them working. Measure the time saved. Then think about the next thing.

Want help figuring out what fits your company? Get in touch.