Knowledge base

What should you let AI do, and what not?

In short, give AI the work that comes back often and follows fixed rules, let it speed you up on heavy research, steer it yourself on creative or strategic work, and keep the human work (judgment, relationships and responsibility) with a person. AI speeds up someone who understands the problem. It doesn't replace them.

The short answer.

Whether you use AI matters less than what you use it for. It pays to sort your work along two axes, how heavy it is for you and how well a computer can handle it. That gives you four kinds of work, each with its own approach. Misjudging them is where money and trust leak away. People automate something creative that loses its nuance, or they keep doing work themselves that a computer has long done better and faster.

My starting point is down to earth. The gain comes from the person who understands the problem, and the tool alone doesn't deliver it. So I never start with a tool. I start by asking which work is costing you time, money or visibility right now, and only then do we decide together what can go to AI and what belongs with a person. That's exactly what happens in the free process review.

The four kinds of work, side by side.

This is how I divide it, with the approach and a few examples for each kind. The first column is the category, and next to it is what it is and when work falls into it.

Kind of workWhat it isExamplesApproach
Give it to AIEasy for you, easy for a computer (boring and repetitive)Summarizing meeting notes, sorting email, pre-sorting reviews, compiling reportsAutomate, as long as it recurs often and pays for itself
Let AI speed you upHard for you, strong for a computer (digging through lots of data or text)Market research, reading through contracts or campaigns, finding patterns in numbersYou give context and choose, AI does the heavy middle, you polish
Direct AIHard for both (creative or strategic work with nuance)Design and variations, a first prototype, mapping out an approachA person directs. AI provides options, you set taste and direction
Keep it with a personEasy for you, hard for a computer (it's all about you)Leadership, negotiating, winning trust, the first introductory meetingNo AI. This is what makes you irreplaceable

When do you really give something to AI?

For the first category, the boring repetitive work, I use three simple questions. Does it come back at least weekly? Does it always have the same, predictable input and output? And does automating it actually save time or money? Only when the answer is yes three times is building worth it. Otherwise you can easily sink your time into automating something you barely do, and that's exactly the kind of project that never pays for itself.

Watch the second and third categories too, because that's where the trap is. When AI speeds you up or you direct it, you don't let go. On heavy research you provide the context and the conclusion, and on creative work you stay the director. That keeps quality high and stops a system from convincingly producing something that's wrong, which is called a hallucination.

Here's what that looks like in practice.

At Inge de Bruijn Zwembad you can see all four. Pre-sorting 648 reviews and compiling the daily briefing at 07:00 went to AI, because that's repetitive work with fixed rules. Combing through 6,000 checks every 24 hours is speeding up. The system finds the anomaly, and a person decides. Working out what the platform should look like was directing, with me as the director. And signing off on what goes to customers, plus every conversation with management, stayed with a person. That last part is no coincidence. Working with a builder instead of an agency means you talk to a person about what matters.

Frequently asked questions.

Which work is best to automate first?

Start with the boring repetitive work that comes back at least weekly, always has the same input and output, and actually saves time or money when automated. Think sorting email, retyping data, compiling reports. That's the safest first step and the fastest to pay for itself.

Which work should you not leave to AI?

Work that revolves around you, like leadership, negotiating, winning trust and personal contact with customers. A computer can't convincingly imitate that, and it's exactly what sets you apart. Always keep a person in there.

Will AI replace my people?

Usually not. For most small and midsize businesses, AI takes over the boring repetitive work and the heavy research, so your people have time left for the work where they do make the difference. AI speeds up someone who understands the problem. It doesn't replace them.

How do I know what AI can do in my business?

I figure that out with you in the free process review. We go through your week, sort the work into these four kinds, and calculate for each opportunity whether automating it pays for itself. You get it on paper, even if my advice is to keep something with a person.

Curious what you could hand to AI?

Walk me through one process. In half an hour we sort your work into these four kinds, and I point out what can be automated, what AI can speed up, and what you're better off keeping yourself. Honest advice, on paper, to take with you.

No newsletter spam. At most one email a month, and unsubscribing is just a reply.

Walk me through one process