Disclaimer: yes, I used AI to help write this article about knowing when to use AI. That’s not a contradiction, that’s the whole point. The 80% is mine. The editing is not (20%). Walk the walk.

I’m done with the two camps.

Camp one is terrified of AI. Camp two thinks it’s the answer to everything
Like those are the only two seats in the room. They’re not. And frankly, thinking in those terms is quite narrow.

The real question was never “AI or no AI.” It’s this:

What internal, painful, repetitive operation can I automate and where does it actually matter that a human is doing it?

That’s it. That’s the whole game.

Here’s the trap I predict people walking into.

You automate everything. You cut every cost you can find. Your margins look incredible on a slide…. and then the silent bill will arrive.

Your product feels hollow, your support feels like a maze, your brand feels like every other brand and it starts to crack. Not because the technology failed.
Because you forgot something that was never going anywhere:

People love people.

No matter how good the models get. No matter what happens next. That doesn’t leave. It can’t. It’s not a preference, it’s HUMAN wiring.

The 80/20 sweet spot

So what’s the actual answer? Not “automate everything.” Not “reject automation out of principle.” The answer is a ratio. This was actually something my dietitian told me about how to obtain the best long term results on my body (true story).

I’d put it around 80% automated, 20% human and emotional. Automate the painful, repetitive, invisible stuff, the stuff nobody was ever enjoying anyway. Protect the 20% that’s actually the point: the moment someone feels seen, catered to… where your brand created love, laughter, memories.

Think of it like a burger. You don’t eat at McDonald’s every single day your body would revolt. But once in a while? It’s pure pleasure. Indulgence works because it’s rare, not because it’s constant.

Humanity in business is going to work the same way.
Not everywhere, all the time that’s not scalable and it’s not even desirable. But somewhere, deliberately, on purpose. That’s what will separate brands people trust from brands people merely use.

The reversal is already happening

You don’t have to take my word for it. Look at who’s already backpedaling.

Klarna built its whole 2024 narrative around AI: its CEO said the company’s AI assistant was doing the work of 700 customer service agents.
=> A year later, that same CEO admitted publicly they’d gone too far, that quality dropped, and started hiring humans back into the loop.

Trade Republic sure seems like they’ve been on the crisis com room and have now launched a “human driven” product launch… and “human committee council”.

Air Canada got taken to a tribunal and lost after its chatbot invented a discount policy that didn’t exist. The airline argued the bot was “a separate legal entity” responsible for its own words.
=> The tribunal was not amused.

Duolingo leaked an internal “AI-first” memo, got dragged publicly for weeks, and had to walk the messaging back in a hurry.
=> Dam Suzan we said we’d switch off the note taker for this one!!!

None of these are anti-AI stories.
They’re companies who ran the automation math without running the humanity math, and paid for it in trust, not just headlines!

What the data is starting to say

Gartner predicted that over 40% of agentic AI projects will be canceled by 2027. Here’s the part that should actually worry you: they named the reasons, and “the AI wasn’t smart enough” isn’t one of them.

The three causes are escalating costs, unclear business value, and inadequate risk controls. Not a model problem. A judgment problem.

But the detail that stopped me is this one: of the thousands of companies out there claiming agentic AI capabilities, Gartner estimated only about 130 were building anything that actually deserved the label. The rest were chatbots and automation scripts wearing a new jacket. And it’s not staying in the pilot phase quietly, either. Forrester found that roughly three-quarters of enterprises have adopted agentic AI, but only a sliver of them are actually running it in production.

Basically a fancy way of saying: it performed for the room, then it met reality.

Meanwhile, the parts of these systems that take real action: sending the email, moving the money, not just suggesting it: jumped from a quarter of all agent activity to nearly two-thirds in sixteen months. Autonomy is scaling faster than the guardrails around it.

Here’s my prediction, and I’ll happily be held to it:

“Human-driven” is going to become a certified claim. The same way “organic,” “cruelty-free,” and “carbon neutral” became labels people look for because the default drifted so far from the thing itself that the exception needed a name.

We’re going to see it on landing pages. In pitch decks. In brand guidelines. Not as a soft value statement, but as a hard differentiator, because the market will be flooded with the alternative.

Why everyone looks the same

And here’s the part that actually makes me laugh.

Scroll through enough websites, enough decks, enough “innovative” brands right now, and you’ll notice something: they all look identical. Same fonts. Same gradients. Same three-word taglines. Same tone of voice that reads like it was run through the same fifteen templates because, increasingly, it was.

Being a contrarian is scary. It’s genuinely not the safe move. It’s much easier to copy what’s already working, ship it, and move on. I get why people do it.

But I keep wondering: when did everyone stop thinking before running?

The brands that will win the next decade aren’t the ones that automated the fastest. They’re the ones that knew exactly where not to.

That’s not a fear of AI. That’s just good judgment.