If AI had a CV, it’d sound like a show off.
It writes emails, runs chatbots, takes meeting notes, flips burgers and, apparently, tops ice cream with bacon.

Yes, that really happened.

McDonald’s rolled out ‘AI-powered’ drive-throughs to make life easier for staff. Instead, one customer drove off with $211 worth of chicken nuggets they didn’t order. Another got a sundae with a porky surprise. The pilot quietly vanished from over 100 locations.

Taco Bell tried it too. Their CTO, Dane Mathews, shrugged, “Sometimes it lets me down.” That’s one way of putting it. In one clip, a customer seemingly crashed the system by ordering 18,000 water cups. In another, a man got increasingly furious as the AI kept insisting he add more drinks, even after he’d said no.

It might sound like comedy. But this is what happens when companies rush to plug AI into frontline roles without stopping to ask how it fits. And while they may raise a chuckle online, they annoy the hell out of customers, kill trust and leave teams mopping up the mess.

AI was sold as a helping hand, however it feels more like another job to do.

Deployment delusion

AI might be everywhere, but it’s not working everywhere.

According to the MLQ State of AI in Business 2025, 95% of AI pilots fail to deliver measurable ROI. AI has potential, but only when it’s matched to the right problems, not thrown at everything that moves.

Too many businesses fall into what we call “deployment delusion”: the belief that rolling out more AI will somehow make everything better, faster, cheaper. Spoiler — it won’t!

We’ve seen it happen:

• Tasks get automated and no one fully understands why, so they end up automating the wrong thing. Agents stuck fixing bad information instead of helping customers.

Chat tools so stubborn that customers end up begging to speak to a person. So why is the answer always more AI?

The real issue isn’t the tech. It’s the lack of thought behind it.. When AI is deployed without strategy, empathy, or understanding, it doesn’t ‘streamline’ anything; it hijacks the experience.

Hallucinations aren’t harmless

The tech industry loves jargon, and “AI Hallucinations” is the latest term for the errors made by chatbots. People talk about it like it’s nothing serious. But it means AI’s coming out with things that simply aren’t true and doing it like it knows best. And some argue they are getting worse.

It’s how you end up with:

• Fast food orders from hell (possible TV series?)
• Meetings summarised with decisions no one made
• Doctors receiving medical records with the patient’s name swapped
• Staff chasing phantom appointments that don’t exist

And who picks up the pieces? The people AI was supposed to help. Doesn’t make much sense, does it. And the most alarming part is that the AI has no clue it’s wrong. That’s the danger. It can’t exactly raise its hand or ask for help.

Jason Roos, CEO of CCaaS provider Cirrus says, “It always blows my mind how fast companies race toward AI without fixing the basics. Klarna went all-in on automation. The numbers looked good, but customers still wanted to talk to a human. Just goes to show that rolling something out doesn’t mean it’s working.”

Don’t get ahead of yourself

Too many people rush into AI thinking speed equals success. The ones who get it right take their time to understand the people doing the work and the problems they’re trying to fix — before adding tech. The companies making real strides with AI aren’t shouting about ‘disruption’ and ‘transformation’, or chasing the latest bit of tech..

They’re asking:
What’s slowing our people down?
What’s winding our customers up?
What could we make easier, right now?

They’re building systems where AI backs their people, not replaces them. Where automation handles the repetitive grind, so agents can do what they’re there for: thinking, helping, building trust.

Here’s the hard truth for the tech optimists: AI isn’t magic. It’s muscle. And it only works when it’s trained, embedded and aligned with your biggest asset, your team. It mostly fails when it’s forced or focused on the wrong goals.

If your current plan sounds like:

“Let’s automate everything and hire fewer people,”
You’re not innovating, you’re inviting a Klarna-style problem. So if you’re serious about getting it right, here’s where to begin.

Start with a better question:

Where does our team need support?

Then:

• Start small. One pain point. One fix. Prove it.
• Work with your people. If they don’t trust it, they won’t use it.
• Design for flow. Not transformation. Just things that work.
• Measure what matters. Minutes saved. Stress avoided. Trust earned.

Roos goes on to say, “One of the biggest blind spots is readiness. Companies jump straight to implementation without first checking whether their systems, data, or people are even prepared. Half the battle is knowing where to begin. AI should be about progress, not theatre. It’s not the race. It’s how you get there.”

Get AI wrong, and it hurts people first and your reputation right after. So don’t be the one handing out bacon sundaes and wondering why things feel off.
Be the one who listens, pilots, learns and builds something that just works.