The importance of choosing the right AI

"Salim, which one do you use?"

Claude? ChatGPT? Gemini? Grok? Mistral?

It's one of the questions you ask me the most.

In replies to these emails, in messages, in comments, at the end of meetings.

And behind it, there's almost always the same worry.

That you picked the wrong one. And that you're wasting your time with it.

So I wanted to give it a whole issue.

To give you my take, laid out clearly, once and for all.

And I hope that by the end, you'll have a clear answer.

I'll start with the conclusion, because it should be a relief.

For most of us, which AI you choose barely matters.

Here's why. And more importantly, what actually matters instead.

The go-kart and the Formula 1

Picture a Formula 1 driver.

You put them behind the wheel of two different race cars.

They'll feel the difference.

A tenth of a second in a corner. How the engine responds. Braking a touch later.

That's their job. They live in those details.

Now put me behind the wheel of those same two cars.

I won't feel a thing.

To me, they'll both be incredibly fast cars. That's it.

But put me in a rental go-kart, then in a Formula 1.

Now I feel the difference. Instantly.

AI works exactly the same way.

The engineers who test these models all day can feel the difference between two top AIs.

They're right to talk about it. It's their job.

But to write an email, prepare a meeting, summarise a report or think through a decision?

Today's leading models are all Formula 1s.

The only difference you'll really feel is the one between a go-kart and an F1.

The one rule: compare F1s with F1s

There is one trap, though.

And plenty of people fall into it.

Every brand offers several levels.

A fast, lightweight version, often free.

And a more powerful one, the kind that thinks longer before answering.

When someone tells me "I tried that brand's AI, it's useless", I almost always ask which version they used.

And very often, they compared one brand's go-kart with another brand's F1.

So if you compare, compare like with like.

The most advanced model from one against the most advanced model from the other.

You'll see the gap is small.

And that it changes almost every month.

What really makes the difference: what's around it

If I had to choose between two AIs today, I'd barely look at how smart the model is.

I'd look at what's around it.

Because an AI is no longer just a window where you type a question.

The big tools now come with features that completely change how you work.

Projects. You drop in your documents, your instructions, your context, once. The AI keeps them in mind in every conversation.

Workspaces. The AI no longer just answers with text. It builds a document, a table or a small app right in front of you, and you can edit it.

Assistants. You set them up once. One to review your proposals, another to prepare your meetings.

Automation. AI can now act: chain steps together, work on your files.

That's where the gap between tools really opens up.

Far more in these features than in the intelligence of the model itself.

A word on Mistral

I get asked about it a lot.

And I want to be honest.

I love what Mistral stands for.

A European company, ambitious, moving fast. I genuinely want it to succeed.

But today, in my view, the gap is still there.

Both in the models and in the features around them.

To stick with my metaphor, it's more of a Formula 4.

Great for writing, rewording, summarising.

More limited as soon as you want AI to act, chain tasks together, work on your files.

That said, if you prefer Mistral, use Mistral.

A well-driven F4 is already phenomenal next to someone who doesn't know how to use AI.

Or doesn't use it at all.

The boots and the game

I play football.

If I want to get better, I have two options.

I can buy new boots.

Or I can work on my technique, my fitness, how I read the game.

The boots help a little.

The rest changes everything.

Choosing an AI is the boots. Knowing how to use it is the game.

That's also why, in this newsletter, I almost never compare brands.

And why I don't show you the trendy new tool every week.

My bet is that you don't need to pay for a thousand tools.

You need to master one.

To know what it does well. Where it gets things wrong.

And above all, at what point in your day it's worth opening.

And when it's better not to open it at all.

No model will give you those instincts.

They come from using it, day after day.

What I'd do this week

1. Pick one.

The one you already have. The one your company allows. The one whose interface you like.

It doesn't matter. Just take the advanced version if you can.

2. Stick with it for a month.

No comparing. No new tools. Just one.

3. Open it every time you hesitate.

Before a tricky email. Before a meeting. In front of a document that's too long.

Even if, in the end, you decide not to use it.

4. Start by asking it what it can do for you.

Here's the prompt I suggest to get going:

I'm a [your job]. Here's what my typical week looks like:
[describe your main tasks in a few lines].

1. Suggest the 5 tasks in my week where you could help me
   right now, and explain how for each one.
2. Also tell me the 2 tasks I'd be better off not handing
   to you, and why.
3. For the first task on your list, ask me the questions
   you need to help me well, one at a time.

Point 2 is the most important.

An AI that tells you honestly where it isn't useful is the start of a good working relationship.

A month from now, you won't be asking which AI to choose.

You'll know exactly when to use it.

If you remember one thing

If there's one thing I'd like you to take from this email, it's this.

You're not behind.

I know that's not how it feels.

Social media, the news, the posts with four thousand likes all repeat the same thing.

That AI is going to replace everyone. That others are already so far ahead you'll never catch up.

So let's look at the numbers.

In the United States, only 15% of employees use AI at work every day. Nearly one in two never use it (Gallup, May 2026).

In Europe, it's even lower. In 2025, only 15% of people aged 16 to 74 used it for work (Eurostat).

Fifteen in a hundred.

So if you're reading this, if you open your AI now and then, if you're wondering how to use it well…

You're already ahead.

Not of everyone, of course. But of far more people than you think.

And to stay ahead, you don't need the best AI on the market.

You need to keep learning.

To stay curious. To try things. To get it wrong sometimes, and start again.

Do that, and you'll be fine.

And don't forget one thing.

You're unique.

The way you talk to your clients, how you see your work, everything you've learned over the years: no AI has that.

AI won't replace you if you use it to bring out what's most special about you.

That's exactly what we try to do together, every week, in this email.

See you next week.

Salim

Every week, a work situation like this one in The Ergon Brief, the free newsletter. Subscribe to get the next issue.