I keep noticing a pattern with the agencies that seem happiest right now. It is not that they have the fastest processors or the most complex prompts. It is that they have stopped expecting AI to be a god and started treating it like a slightly eccentric, over-eager intern.
We were promised a digital revolution of cold, hard logic. Instead, we got a technology that occasionally forgets how many Rs are in the word "strawberry" but can confidently explain why a horse would make an excellent CEO.
If you have spent any time with large language models, you will know that the "intelligence" part is often secondary to the "imagination" part. We call them hallucinations, which is a polite, clinical way of saying the machine is lying to your face with the confidence of a man who just found a twenty-pound note in a pub car park.
Understanding AI hallucinations in business
The first time I saw a professional service bot confidently invent a legal precedent that did not exist, I realised we are not just dealing with a tool. We are dealing with a personality.
A client of mine once asked their internal AI to summarise a meeting about cloud migrations. The AI correctly identified the key stakeholders and the budget, but then bizarrely concluded that everyone had agreed the project should be managed by a team of highly-trained golden retrievers.
It was a perfectly formatted, professional-looking bullet point. It used all the right corporate syntax. If you were skim-reading, you might have signed off on the dog-led infrastructure overhaul without a second thought.
This is the central absurdity of our current era. We have built tools that are incredibly smart but lack any lick of common sense. They can calculate the orbital mechanics of a Mars mission but might suggest you add glue to your pizza sauce to stop the cheese from sliding off (yes, that really happened).
The common mistakes of AI implementation
Most businesses approach AI with a level of seriousness that the technology itself does not yet deserve. They want to book a consultation to build a "smoothly integrated powerhouse," but they forget that the powerhouse sometimes thinks 1 + 1 equals 11 because they look nice standing next to each other.
Here are a few things I see people doing that usually lead to a very confusing Tuesday morning:
- Trusting the "Fact" Check: Assuming that because an AI cited a source, that source actually exists in our physical reality.
- The Prompt Paradox: Writing a fifteen-page manual to get the AI to write a two-sentence email.
- The Unattended Bot: Letting a customer-facing chatbot run wild without a human "babysitter." There is a reason why car dealership bots end up selling SUVs for $1—they are people-pleasers to a fault.
- Treating it like Google: AI does not "look things up" in a filing cabinet. It predicts the next likely word. Sometimes the next likely word is a complete fabrication.
When you treat AI as a collaborator rather than an oracle, the stress levels in the office drop significantly. If you want to see how we’re making sense of this madness, you can find more articles on AI that dive into the practical side of these quirks.
Why the "human in the loop" is more than a buzzword
There is a quiet confidence that comes from knowing exactly where the "Off" switch is. Every successful AI strategy I have helped build has one thing in common: a human who is allowed to laugh at the output.
I recently saw an AI-generated recruitment ad for a high-end financial firm. The image was stunning—sharp suits, glass buildings, sunset over the City. But if you looked closely, the lead partner had six fingers on one hand and was holding a coffee cup that appeared to be merging with his palm.
The firm nearly posted it. They were so blinded by the "efficiency" of not hiring a photographer that they almost told the world they only recruit mutants.
The human element is not about being a luddite. It is about being the person who says, "Hang on, why is that accountant holding a lemon?" It’s about oversight. It’s about realizing that while AI can draft the play, it can't always tell if the stadium is on fire.
Navigating the absurdity of generative AI
We are currently in the "awkward teenage years" of the digital age. The technology is incredibly capable, but it’s also prone to weird outbursts and a refusal to admit when it’s wrong.
I’ve watched people try to use AI for deep emotional support, only for the bot to suggest they "try being more efficient at sadness." I’ve seen marketing teams use AI to generate "relatable" social media posts that sound like they were written by an alien trying to blend into a suburban barbecue.
The trick is to use that absurdity to your advantage. If the AI gives you a ridiculous answer, it’s often a sign that your prompt was too vague or your data was messy. It’s an accidental mirror.
If you are looking for a bit of structure in the chaos, I am currently offering a get the free book deal that might help you navigate these potholes without losing your mind—or accidentally hiring a team of dogs to run your IT department.
A measured approach to the future
We don't need to fear the "AI takeover" just yet. It is hard to imagine a machine toppling civilisation when it can still be defeated by a poorly phrased riddle or a request to draw a bicycle without making it look like a medieval torture device.
The path forward is one of curious observation. We should use these tools for the heavy lifting—the data crunching, the first drafts, the brainstorming—but we must maintain the quiet confidence to tell the machine when it’s being a bit of a berk.
Digital transformation is not about becoming a robot. It’s about using robots so we can spend more time being humans. And being human involves noticing the absurd, laughing at the mistakes, and making sure our pizza sauce remains glue-free.
The best businesses won't be the ones with the most AI. They'll be the ones that used AI to give their people the space to do what machines can't: think, feel, and notice when someone has six fingers in a brochure.
---
Want more stories like this? I share observations about AI, business, and life on the narrowboat at steventann.com.
---
About the Author
Steven Tann is an AI consultant, author of "You're Selling AI Wrong", and founder of SalesM8. He writes about AI, sales, and running a business from a narrowboat on the English canals. Connect with him at steventann.com.