I Sell AI for a Living.
I’ve Never Sold One That Replaces Your Staff.

The one claim I have never made

I have shipped AI into real businesses. Identity, commerce, a generative media studio, chat assistants that handle live customers. Real products, real invoices, real Malaysian companies depending on them.

Not one of them is sold with the line “fire your staff.” That is deliberate, and I want to explain why — because the reason is the opposite of what you would expect from someone whose business is AI.

I don’t avoid the replacement pitch because I’m cautious, or noble, or worried about the optics. I avoid it because I deal with what’s actually under the hood every single day — and the people who build the engines are far less confident than the people selling the ads.

The people selling the shovels won’t promise you the gold

I work with the big model providers constantly — pricing, rate limits, what a model will and won’t reliably do in production. And here is the pattern nobody advertises: the providers hedge on autonomy. Their documentation is full of “human-in-the-loop,” “review before acting,” “not suitable for high-stakes decisions without oversight.” The companies with the most advanced AI on earth put a human checkpoint in their own instructions.

You don’t even have to take my word for it. Look at how the most AI-forward companies run AI on themselves. IBM automated the routine work of its own HR department with an agent called AskHR. It now settles about 94% of routine staff requests. Impressive — and the number everyone quotes.

But read what IBM did with the other 6%. The remaining cases — the ethical calls, the exceptions, the ones that need judgment — still go to people. And then IBM went further: it announced it would triple its US entry-level hiring for 2026. Its own HR chief, Nickle LaMoreaux, said the quiet part out loud — if you stop hiring junior people because AI covers the easy work, “in three to five years there’s no pipeline. The well simply dries up.”

The most sophisticated AI operator in the room automated the boring 94% and spent the savings hiring more humans. That is not the behaviour of a company that believes AI replaces people. That is a company that knows exactly where AI stops.

The receipts are already coming in

While the ads keep promising, the companies that believed them are quietly walking it back.

Ford rehired around 350 veteran engineers after discovering that automated quality systems couldn’t catch problems that experienced people catch on instinct. Its own hardware engineering VP put it plainly: AI is a good tool, but its output is only as good as the data it’s trained on — it can’t replace human engineering judgment. Ford then went on to top an independent 2026 quality study. The lesson wasn’t “AI failed.” It was “we removed the people who made AI work.”

Commonwealth Bank of Australia cut 45 customer-service roles, saying an AI voice bot had reduced call volumes. The union pointed out that calls were actually rising and staff were being pulled into overtime to cope. In August 2025 the bank reversed the redundancies and admitted its assessment “did not adequately consider all relevant business considerations.” Translation: it fired people on a number that looked good in a slide and fell apart in reality.

And it isn’t two anecdotes. A survey of a thousand senior leaders by the workforce platform Orgvue found 39% had made redundancies because of AI — and 55% of those admitted the decision was wrong. Separate reporting put roughly a third of companies already rehiring for roles they’d automated away. More than half of the executives who fired people for AI now concede they shouldn’t have.

So where does the confidence come from?

This is the part that genuinely puzzles me. I have the most direct possible view of what these systems can do — I pay for them, I test them, I ship them, I get the 3am support ticket when one misbehaves. And I would not stake a person’s job on a claim of full replacement. Yet CEOs with none of that exposure do it constantly.

I’ve come to think the confidence has a specific, unglamorous source: the ad, not the understanding. An old-school CEO doesn’t read the model card or run the pilot for six months. He sees a slick LinkedIn ad — “replace your whole support team, 90% cheaper” — forwards it to his CFO, and the headcount conversation writes itself. The confidence isn’t earned from the technology. It’s borrowed from a marketing video made by someone who has never had to run the thing in production.

There is a real gap between “AI can do this” in a demo and “AI can do this” in a business at 3am with a real customer and real money on the line. The demo is where the ad lives. The 3am is where I live. The distance between those two is the entire job — and it’s exactly the distance the replacement pitch pretends doesn’t exist.

What I build instead

Everything I ship is built to make a person do the work of three — not to remove the person. The chat assistant qualifies leads so a human closes better ones. The media studio lets one marketer produce what used to take a team. The identity and commerce layers remove the drudgery so people spend their hours on judgment, relationships, and the exceptions machines are bad at.

I use one test on every product before it ships: when the AI is wrong at 3am, who catches it? If the honest answer is “no one, because we fired them,” then you didn’t automate a process — you gambled on one. Automation with a human on the exceptions is leverage. Automation with nobody watching is just risk you haven’t been billed for yet.

Where AI genuinely does replace work — because I won’t pretend otherwise

I’m not going to tell you AI replaces nothing. That would be its own kind of dishonest ad. AI genuinely removes tasks — and where a job is almost entirely one repetitive task, some real displacement follows, and it will keep happening. Anyone who tells you no jobs will change is selling you comfort.

But “AI removes tasks” and “AI removes the human” are different claims, and the gap between them is where careers and companies get wrecked. The honest version is: AI changes what a person spends their day doing. The dishonest version — the one on the ads — is that you can keep the output and delete the person. The companies quietly rehiring this year are the ones who believed the dishonest version and got the invoice.

The bottom line

The confidence to replace a human being should come from running the system yourself, in your own business, for long enough to see where it breaks. It should not come from a video made by a vendor who has never carried your risk.

I run these systems. That’s precisely why I’ve never sold you a product that promises to replace your people — and why I’d be a little suspicious of anyone who does.

Published by IMA AI — July 2026. We build AI that makes people harder to replace, not easier to fire — which is why none of our products claim otherwise.