← Back to Blog

The AI Sales Pitch That Keeps Blowing Up in Everyone's Faces

Klarna sacked its human agents for AI, then had to hire them back. Apple paid $250 million over an AI feature that didn't exist yet. The problem wasn't the tech. It was how AI got sold.

By · · blog

The AI Sales Pitch That Keeps Blowing Up in Everyone's Faces

The AI Sales Pitch That Keeps Blowing Up in Everyone's Faces

I live on a narrowboat. Slowest form of transport in Britain, probably. You cruise at four miles an hour, you queue for locks, and every so often you have to get out and wind a paddle by hand because that's how it's worked since the 1770s.

People assume that means I'm anti-tech. Mate, I run an entire business off this boat. AI drafts my emails, chases my leads, writes half this blog, and posts to social media while I'm asleep. I'm about as automated as a one-man operation gets.

But here's the thing four miles an hour teaches you that the AI sales pitch never does: you still have to steer the boat. Nobody's selling that bit. Everyone's selling the bit where you sit back and the boat drives itself to the destination while you have a cup of tea.

That's the wrong pitch. And it's costing some very big companies a lot of money to find out.

Klarna sacked the humans, then had to ring them back

Sebastian Siemiatkowski, CEO of the buy-now-pay-later firm Klarna, spent 2023 and 2024 telling anyone who'd listen that AI could do the jobs of thousands of people. Klarna slashed its outsourced customer service headcount by roughly 2,000 and stopped hiring altogether. Siemiatkowski said he wanted Klarna to be OpenAI's "favourite guinea pig."

Eighteen months later, Klarna reversed course. The company started hiring human agents again in 2025 so customers could always reach a real person. Siemiatkowski told Bloomberg the AI replacement had produced "lower quality" service. A Silicon Republic report at the time noted Klarna hadn't hired a single human throughout 2024, right up until the U-turn.

Klarna's not a small business run by a bloke and a laptop. It's a fintech giant with a research budget most of us couldn't dream of. If they got the pitch wrong at that scale, that tells you something about the pitch, not just the company.

Apple paid $250 million because the pitch got ahead of the product

In 2024 Apple marketed a "more personalised Siri" as a headline reason to buy the new iPhone. The trouble is the feature didn't exist yet. It still doesn't, fully. Buyers sued, arguing Apple oversold AI capability that wasn't real, and in May 2026 Apple agreed to a $250 million settlement to close the case. The Verge reported the payout could put up to $95 back in the pocket of every affected buyer.

That's not a scrappy AI wrapper company overpromising in a sales deck. That's Apple, the most valuable company on the planet, getting the sell wrong badly enough to write a nine-figure cheque for it.

The pattern behind both stories

Neither of these were technology failures in the way people assume. AI didn't "not work." Klarna's AI could still answer a lot of customer queries. Siri could still do plenty of useful things. The failure was in the pitch: sell the fantasy of full replacement or magic new capability, then have reality turn up and demand receipts.

And it's not just the giants. MIT's NANDA initiative studied enterprise AI pilots and found 95% delivered no measurable return to the bottom line. Not "didn't work at all." No measurable return. The tools were there. The rollout was sold as bigger than it should have been, deployed without the boring bits (training, workflow redesign, someone actually owning the outcome), and the result was an expensive nothing.

I see the same pattern in miniature every week in the GHL world. An agency sells a client "an AI receptionist that handles everything." Three weeks later the AI has told a customer something untrue, or missed an obvious "call me back urgently," and the agency owner is doing damage control on a Friday afternoon. The tech wasn't the problem. The pitch was.

What selling AI properly actually looks like

I've automated a genuinely large chunk of my own business from a boat with patchy 4G. Here's what I've learned about the pitch that actually holds up, versus the one that doesn't.

Sell the boring bit, not the destination. Nobody buys "AI that runs your whole business." They buy "AI that drafts your inbox replies so you only touch the ten that matter." Narrow, specific, provable. The moment the promise is "everything," you've already oversold it.

Somebody still has to steer. On the boat, the engine does the work, I decide the direction. In the business, AI drafts the content, chases the lead, flags the invoice. I still decide what gets sent, who gets called back personally, and what doesn't sit right. Sell that division of labour honestly and clients trust you more, not less.

Measure the actual thing, not the vibe. "AI will transform your operations" isn't measurable. "This cuts your inbox triage from three hours a week to twenty minutes" is. If you can't point to the specific hours or pounds it saves, you're selling hype, and hype has a habit of turning into a Friday afternoon apology call.

Small and proven beats big and promised. Every AI rollout I've seen actually stick started with one task done properly, not five tools bolted on at once. Klarna and Apple both had the resources to do this right and still got burned reaching for the bigger promise. A five-person business has even less room for that mistake.

If you want a proper starting point rather than a vague AI ambition, the free AI Setup Kit is built around exactly this: narrow, specific tasks with markdown skills you can actually run, not a grand plan that needs a consultant to translate it.

The bit nobody puts in the ad

Living slow on a boat and running an automated business at the same time isn't a contradiction once you see what AI is actually good at. It's good at the repeatable, the tedious, the three-in-the-morning admin task. It's rubbish at replacing judgement, context, and the decision about who gets a real phone call today.

Sell that, and the AI genuinely earns its keep. Sell the fantasy, and sooner or later you're the one writing the reversal statement, or the settlement cheque.

Sources:

Want to try this yourself? Grab the free AI Setup Kit. 77 plug-and-play skills for content, sales, marketing, and operations. Works with Viktor, Claude, or ChatGPT. No course, no tool to configure, just markdown files and your AI.

Brewed by Steven, poured by Viktor

About Steven Tann: Steven helps business owners build systems that run themselves using AI. After 10+ years helping 7,000+ businesses and building his own autonomous operations, he's the bloke who actually does it, not just talks about it. Find out more at steventann.com.

Tags: Small Business Automation, AI for Small Business, Practical AI, AI Failures, The Autonomous Business