Bringing value now matters more than ever

Every decision started with a deep dive

I've worked in quite different places: selling wine, running a hotel, managing construction sites, and now building software. The industries have little in common, but one thing kept coming back. Every real decision needed a decent deep dive into the topic first.

Before you can say yes or no to a new supplier, a renovation, or a new feature, you need to understand what you're actually agreeing to. That means reading, asking, comparing, and sometimes a few evenings of doubt.

Saying "no"

The deep dive mostly protected me from one thing: saying yes to everything. Every yes has a price, in money, time, or attention, and it's easy to underestimate all three.

Looking back, about 90% of the times I said no, it was the right call. Not because the ideas were bad, but because they didn't bring enough value for what they would cost.

I'll admit it, though: learning to say no was extremely hard, and not only at work. I'm still working on it.

Everything is deliverable

Once AI entered the game, the situation changed drastically. Suddenly everything is deliverable:

• a big, fancy presentation for a client? not a problem
• a 3D visualisation of a building? not a problem
• a new feature in the application? not a problem

All of that, and an infinite number of other things, can be done in minutes, hours, or days. Producing things has become cheap. What comes after hasn't: someone still has to check it, use it, and maintain it. Quite often, that someone isn't you.

The problem with "everything is possible"

I won't go into the general use of AI here. There are plenty of articles about that already. I want to focus on one thing I see as one of the biggest future problems for all of us: letting AI decide everything for us.

When everything is possible, the hard part is no longer doing the thing. The hard part is deciding whether the thing should exist at all.

We used to own the conclusion

In the past, it was always up to us what we took from the research, the deep dive, or whatever we call the preparation phase, and how we used it. The conclusion was ours, and so was the responsibility.

Now there's something that offers to save us all the hard work of coming to a decision. And let's be honest, we all like it. Less reading, less thinking, more action. Who wouldn't sign up for that?

The catch is that the thinking was never just overhead. It was where we learned, and where we built the judgement for the next decision.

Where is the value?

You've probably already seen both the positive and the negative impact of working with LLMs, especially in teams. It's hard to keep up:

• tons of documents to read
• endless presentation slides to go through
• an unbelievable amount of code to review

Very often the problem comes from the other side of the conversation. A person takes the AI output exactly as it came out and sends it to you. No revision, no cleanup, just forwarded, as if you had all the time in the world.

This is the moment I worry about the most. Where is the value here? Who created it? What did the person forwarding it learn, and what do you gain on the receiving end? Most of the time, almost nothing.

Isn't it just machine-to-machine communication?

If one side generates the text without reading it, and the other side summarises it with AI because there's no time to read it, then the humans in the middle are just a very expensive message queue.

We still hold the wheel

I'll be perfectly honest with you. The new frontier models deliver great output. Sometimes even better than ours.

But let's not forget that we still hold the wheel, and we need to keep holding it for a while. Otherwise, what's the point of us being here?

The responsibility stays with you

Whatever tool we use, we shouldn't forget who is responsible for the decision. The tool can suggest, draft, and even argue its case well. The decision, and everything that follows from it, is still ours.

Try telling a colleague or a client: "It's not my fault, my agent told me so." In most cases, they won't be happy. Nobody signed a contract with your agent. They work with you, and they trust you to stand behind what you deliver.

If you can't explain why you made a decision, you didn't really make it. You just passed it along.

Can we? Yes. Should we?

When we work on something, it should still be mainly up to us to decide whether it brings real value to anyone or anything.

Does a luxury Rolex need an LED flashlight? I guess not so much, maybe in some edge cases. People buy it for craftsmanship and status, not for finding their keys in the dark. For a sports watch brand, a flashlight is definitely valuable. Can we put one on every watch? Yep. Should we? No.

Summary

AI made producing things cheap. It didn't make deciding, checking, or maintaining them cheap. Before you build, generate, or forward anything, ask yourself:

• who is this for, and what do they gain from it
• did I add anything, or am I just passing it along
• can I explain and stand behind this decision myself
• how would I react if this landed in my hands

The deep dive is still worth it. These days, it's probably worth more than ever.