Why AI Is Destined to Fall Short: A very British response to an artificial problem.
by Julian Reeve
09.01.2026
A single project just convinced me that AI will fall well short of its ambitions. Why? Because something fundamental to its success is completely missing.
Over the past few weeks I’ve been focusing on sales. I needed a new landing page specific to the task, and opted to use ChatGPT and Codex to help me.
Early progress was encouraging. Codex did most of the heavy lifting, reacting to detailed prompts efficiently and effectively. The Codex spend was noticeable, consistently creeping over my intended budget, but I valued what was being produced, so pushed on.
Then, relative disaster struck.
On completion, I noticed the page was loading slowly and didn’t feel stable. I questioned ChatGPT as to the reasoning, and its answer left me speechless.
-
ME: The page feels super clunky and isn’t loading cleanly. What’s the problem?
GPT: I have incorrectly built this page in one single code block, which makes this page extremely heavy. I should have used the software’s drag and drop feature.
ME: Why didn’t you use that feature in the first place?
GPT: The first task required code. I created a single piece for the first task, and incorrectly added to that code each time you asked for something else.
ME: Even though I have continually asked you to build this in the most efficient and responsive way possible?
GPT: Yes
-
The ‘solution’ conversation went like this:
-
GPT: The only real fix is a complete rebuild of the page using the software’s drag and drop feature.
ME: How easy is that and what percentage of the visual quality will I lose?
GPT: It’s a relatively simple task, and I estimate you’ll only lose 10% of the visual quality.
ME: Proceed.
-
I wish I still had the visual proof of what it came back with.
My 12-year old nephew could have done better - 11 years ago!!
Once my blood pressure (and bank balance) had improved, I evaluated what had happened. I went back through my prompts, understanding where I perhaps could have been better with my direction, learning more from the patterns that had unfolded.
Then it hit me.
I realized that I’d questioned GPT on several occasions surrounding specific build choices, only for it to continually make the same mistakes.
It had spent the entire project assuring me that we were working in the right way, only for it to do the complete opposite afterwards.
So, I did what any quietly fuming Brit would do ...
I instructed the bot to write a complaint letter to OpenAI asking for my money back. (If you know me, that won’t surprise you)
What it returned with was by far its strongest work; succinctly arguing the case against itself and its continual failings throughout the project. (It was entirely brilliant! Irony, indeed.)
OpenAI’s response?
“We’re responsible for providing the data, not the results. Not our problem. Request DENIED!’
Sometime later, I understood the deeper lesson.
ChatGPT or any other Gen AI will continue to make these kinds of mistakes because there are no consequences if they don’t. None.
In addition, OpenAI and other AI companies will continue to deny responsibility because they don’t want to be held accountable.
Remove consequence and accountability, what do you get?
Slop.
(The very smelly kind.)
So where does all this leave us?
Do we unleash the full extent of our expletive vocabulary at our computers until AI somehow changes its ways?
Or do we connect with the deeper realization that we may all need to say ‘enough is enough’, and (once again) design our workflows to lean on (human) consequence and accountability for better results?
Whichever path you choose, one thing is clear. AI simply can’t achieve what they say it will without the deeper human capabilities we’re told it will replace.
PS: I’ve since asked ChatGPT to write a response to OpenAI. Its work was outstanding! I’ll keep you posted.
Continue the conversation
This essay is part of The Intentional Creative series, my ongoing exploration of human capability, identity, creativity, judgment, and sustainable performance as technology transforms the way we work.
Subscribe to receive future essays and field notes exploring how we can create unmistakable identity and translate human capability into human advantage.