Implementing AI in your business: where do you start?

Updated: August 2026

AI can do a lot these days, and that is exactly the problem: where do you start? Most businesses start with the technology and then look for a problem to match. The reverse route works better. This is the step-by-step plan we use ourselves.

Step 1: start with the work that comes back every week

Forget for a moment what AI can do. Take a list and write down where your team loses hours every week: answering the same customer questions, retyping requests, drafting quotes, writing up reports, moving data from one system to another. That list is your starting point, not the technology.

Rule of thumb: work that recurs often, runs predictably and consists of reading, writing or transferring information is promising. Work that requires judgement, feeling or responsibility stays human.

Step 2: pick one process, not five

The temptation to tackle everything at once is real. Don't. Pick the one process where the pain is biggest or the gain clearest, and finish that first. One working solution the team uses daily is worth more than five half-finished experiments.

Starting small has another advantage: you learn what working with AI means in your business (what the team thinks of it, where the data needs to come from, which agreements are needed) while the investment is still small.

Step 3: look at your data and systems

AI runs on information. Before anything is built, the question is: where does that information live today? In a tidy system, or spread across mailboxes, Excel and people's heads? That determines how much preparation costs.

Good news: it does not have to be perfect. A working solution on the data you have is often the best starting point for getting your information in order anyway. But know where you stand beforehand, so the quote does not come as a surprise.

Step 4: agree what stays human

The best AI solutions in SMEs prepare work instead of taking it over. The assistant drafts the answer, the automation prepares the document, and a person checks and sends. Agree up front where that line is and what happens when in doubt.

That is not just sensible, it also makes the step smaller for your team. Nobody is replaced; everybody loses the most tedious part of their job.

Step 5: measure whether it works

Agree beforehand how you measure success: fewer hours on a task, faster response time, fewer errors. Check after a month whether the promise holds. That keeps the supplier sharp and yourself honest.

Does it work? Then the next process on your list is up, with everything you learned from the first round.

Frequently asked questions

Do I need technical knowledge in-house?

No. You need to know your own work, that is the knowledge that matters. The technology is the builder's responsibility; you mainly need to be able to say sharply what the result should be.

How long does such a first step take?

A first working solution for one process is usually up within a few weeks. The inventory beforehand, for example with our Work Scan, costs you half a day of your time.

What if it turns out not to work for our business?

That is why you start small and measure the result. A clearly scoped first step can be stopped without a big loss. Be wary of projects that only deliver something after a year.

Want to know where AI starts in your business?

The Work Scan (€495 excluding VAT) maps out which work in your business lends itself to AI and what that yields. Concrete report, clear priority, no obligations afterwards.

Request an introductory call