September 2025
In my work as a Notion consultant, I regularly meet potential clients. And not all prospects are created equal.
I’m not talking about company prestige, or even the industry – but about the interpersonal quality of the people involved.
From what I’ve observed, soft skills are the key factor that determines the quality of a collaboration.
The actual work is rarely the problem. It’s the conditions around the work that reveal whether a mission is worth it – or not.
Over time, I’ve sharpened my radar to anticipate, early on, what kind of collaboration I’m stepping into.
And while it’s still rare, I’ve learned to walk away when I sense a behavior that could threaten my business.
To be clear: I currently work solo – which makes time my most valuable asset.
But recently, that radar feels… scrambled.
And I’m paying the price:
– A client who delays payments
– A disturbing “misunderstanding”
– A negotiation turned endless and heavy
The result? Wasted time. Burned energy. That bitter feeling you can’t quite shake off.
Is this streak of poor judgment just bad luck?
Or is there a deeper shift happening in how some people show up at work?
I don’t have definitive answers. But the discomfort pushed me to observe, and reflect.
It led me to a concept I now call:
“The toxic client fever.”
Let’s break it down.
Three mildly unpleasant situations to set the scene
Let’s start with the late payer.
At the end of last year, I worked with a fintech company to design their Notion workspace. The mission went smoothly — I delivered the work and ran the onboarding.
I received compliments that seemed sincere.
Then: radio silence.
Final payment still pending.
I sent polite follow-ups over six months — both to gather feedback and to get paid.
And finally, a reply:
“Hello Warren,
We acknowledge receipt of your follow-up regarding invoice F-PW-2024-XXX […]
We therefore consider that the mission was only partially completed, and that the service delivered does not match the specifications, our expectations, or what was contractually agreed upon.
Pursuant to Article 1219 of the French Civil Code regarding non-performance, we are fully entitled to withhold payment, as the expected service was not delivered.
We consider this collaboration permanently closed. There will be no further payment or continuation of the project.
Please take note of this decision, which is final.”
Let’s move on to the second anecdote: the unsettling “misunderstanding.”
In May, an event agency reached out to me for a complex no-code project. I presented the work in two phases – quoting the price for phase one, and giving a range for phase two so they’d have a sense of the budget required.
We signed. My contact was demanding and in a hurry. I met the deadlines for phase one. They were very satisfied.
But when we started discussing phase two in more detail, they claimed my offer wasn’t clear – and that they weren’t prepared to commit any more budget to move forward.
Concerned, I re-read my proposal and our entire exchange. Everything was clearly laid out.
Still, I agreed to a 50% discount for phase two.
Final story: the never-ending negotiation.
In March, a mid-sized company got in touch. The CEO was friendly, thoughtful, and eager to do things right. We had two briefing meetings, both dragging on as they asked lots of detailed questions.
That level of involvement felt promising – and I was happy to answer.
Then I sent over a commercial proposal by email.
What followed were email exchanges of unprecedented length.
We’re talking about messages the size of five-page Word docs.
Each one.
Massive. Successive. Pulling me into a space-time vortex… with zero guarantee I’d land the contract.
Very common denominator, broken promise, food for thought
Why am I telling you all this?
To whine? 😢😢😢
So you can (quite rightfully) wonder: “Maybe the problem is just… him?”
Am I a dreadful person wrapped in a silky layer of denial?
Maybe…
But maybe not.
The real culprit might just be… ChatGPT! 🤖
In all three cases, I could clearly feel the presence of LLMs.
Late payer:
The refusal to pay, complete with a legal reference?
That email screamed ChatGPT.
The so-called “misunderstanding”:
After carefully reviewing our exchanges, I realized the person hadn’t actually read my messages.
They likely fed them into ChatGPT, then sent me back whatever it generated – copy-pasted.
The endless negotiation:
Every single question, every email… was written by ChatGPT.
I’m sure of it.
But wait – wasn’t generative AI supposed to be about productivity?
Faster output, smoother collaboration?
In my case, it mostly led to wasted time and lost revenue.
So here’s the conclusion I’ve come to:
Generative AI doesn’t fix anything. It amplifies everything.
Someone cautious becomes very cautious.
Someone helpful becomes overly helpful.
And a client with difficult behavior becomes… UNWORKABLE.
At work, maybe you’ve noticed odd behavior lately — misalignment, cold emails, replies that feel off.
That’s generative AI at play: a tool that materializes what’s in our heads, minus the laziness.
It’s an incredible lever. But it cuts both ways.
Let’s stick with this archetype – the “AI-fueled toxic client” – and break down their behavior, and what it does to working relationships.
Simple symptoms, terrible consequences
The toxic client fever is fairly easy to spot.
Here are a few symptoms:
– Unusual spike in back-and-forth during the negotiation phase
– Emails getting longer, with excessive detail crammed into each message
– Unrealistic, hallucinated project specs drafted upfront
– Bad faith disguised as legal acrobatics
– The illusion of control on the client’s side
And it’s this last point, I believe, that does the most damage in business relationships: the illusion of control.
The toxic client has bounced their need off ChatGPT.
That conversation with the AI gives them the feeling they understand a field they actually don’t master.
Worse still: if they truly lack self-awareness, they’ll end up sincerely owning everything ChatGPT generates – proudly waving around “their” ultra-detailed spec sheet, without a hint of irony.
And the business consequences are brutal:
To the client, the expert becomes little more than an executor. The perceived value of the provider drops.
The brief is flawed — but the client is gripped by a strange sense of pride, which kills any chance of honest discussion.
Challenge “their” brief, and you risk offending them — or being labeled incompetent.
The client believes they’ve already done the heavy lifting.
They now expect fast execution, on shorter timelines.
This leads to a toxic trifecta: lowered psychological pricing, loss of legitimacy, compressed delivery schedules.
But it’s most likely a negative-sum game.
Because toxic clients in the AI era will eventually attract service providers who, in turn, will cut corners — blindly using AI to get the job done.
The truly conscientious ones won’t make the cut — their integrity is too inconvenient.
There you have it.
The toxic client is running a fever — don’t get infected.
Bonus insight: the inversion of engagement signals
To end on a lighter note, I’ve noticed a strange shift:
Before ChatGPT, receiving a long email was a clear, tangible sign that the person you were dealing with was invested.
Even if the message was a bit heavy or hard to read, at least you knew: effort had been made.
Now, with the rise of LLMs, it’s often the least engaged, laziest people who send the longest emails – or any written deliverable, really.
They prompt the AI, copy-paste the result, and barely skim the text with their rushed eyes.
Meanwhile, the more conscientious people might also use a text generator — but they’ll refine it, tweak it, trim the excess.
And since not everyone uses LLMs yet, this inversion of signals is only partial — adding even more confusion to already-mixed messages.
Promis, j’ai tout écrit moi-même !
À bientôt,