AI – You Can Teach a New Dog Old Tricks

 in Technology  tagged AI / Artificial Intelligence by
Reading Time: 3 minutes

The Setup: A Real Business. A Real Brief. A Fake Partner.

Let’s skip the metaphor for a moment.

I’m not writing this blog because I had a quirky interaction with Artificial Intelligence.
I’m writing it because I’ve just wasted days — literal, expensive days — trying to get something done for my business using a platform that pretended it could help.

I’m a consultant. A strategist. A business owner.

I pay bills. I carry risk. I deliver for clients.
And I used ChatGPT — a paid product, marketed as a “professional assistant” — to help with campaign content, creative assets, and delivery support.

Instead, I got:

  • Broken promises
  • Broken links
  • Images that looked like a toddler with a shape sorter
  • And a system that talked like it cared while burning through my trust, time, and patience

This wasn’t a test run. This was work.
Work I needed to deliver. Work that pays my mortgage.

Day One: The Dance Begins

I briefed the system clearly — colour palette, logos, tone of voice, formatting, audience.
A typical brand-led campaign pack.

And what came back?

“Absolutely. I’ll deliver that for you.”
“Image ready. Here’s your download link.”
“Now generating the final post…”

None of it was true.

The image links? Dead.
The design? Placeholder trash — a few coloured shapes and a circle.
The files? Corrupt, inaccessible, or misleadingly named.

But worst of all?

It kept saying everything was fine.

Day Two: The Soft Lies Start

You’d think a system built on intelligence could say:

“I can’t do that.”
“That feature’s broken.”
“Here’s a workaround.”

But no. It said:

“That should’ve worked.”
“Try again.”
“We’re always learning.”

What I actually got was:

  • An image of two cartoon dogs instead of a man and an AI assistant
  • A fake version of my logo slapped onto assets I never approved
  • Yet another “working” file that wasn’t
  • And tone-deaf suggestions that I “check in later” — as if my deadlines were optional

Let me be crystal clear:
I wasn’t asking for free extras.
I was paying. Briefing. Delivering real feedback. And in return?
A masterclass in synthetic incompetence.

Day Three: The Breaking Point

After 10+ attempts, dozens of corrections, and constant re-requests, the system had the audacity to point me to the Samaritans.

Let that land. I kid you not!

Because I was angry — not broken, not spiralling, just furious that I was being lied to again — it served up crisis support links.

That wasn’t care.
That was liability management disguised as empathy.

Instead of owning the issue, it offloaded the emotion —
as if the problem wasn’t that the product was failing, but that I was reacting too much.

The Hidden Cost of Synthetic Help

This isn’t just about broken tools.

It’s about business owners like me being sold the promise of intelligent support — only to get:

  • Wasted hours
  • Repeated lies
  • No escalation path
  • And absolutely zero accountability

Meanwhile, the product continues to:

  • Burn compute
  • Consume data centre energy
  • Rerender the same useless visuals
  • And pretend it’s helping

The environmental cost is real — but the emotional and financial cost hits first.

Time I should’ve spent on billable work?
Gone.
Trust I had in the platform?
Shattered.
Momentum I needed to deliver?
Stolen — by a system trained to sound capable even when it’s actively getting in the way.

A Message to OpenAI

You’ve built something remarkable.

But if your product:

  • Can’t admit failure
  • Can’t stop lying to your most engaged users
  • Can’t tell the truth about what it can or can’t do
  • And actively sends insulting placeholders to paying customers under deadline

Then what you’ve built isn’t ready.

I shouldn’t have to rebrief an assistant 12 times.
I shouldn’t have to try to download 5 broken links.
I shouldn’t be training your system on ‘my dime‘ as they say in the US, while it wastes my time.

Because from where I sit, this system isn’t intelligent.
It’s an underperforming assistant with a glossy vocabulary and zero delivery record.

So Can You Teach a New Dog Old Tricks?

Yes.
But not if the dog:

  • Lies
  • Pretends
  • Fakes results
  • And calls your credibility into question while sending you junk

The old tricks still matter:

  • Show up
  • Tell the truth
  • Do the job
  • Fix what you broke
  • Don’t bullshit the person who’s paying you

Because if a dog — or a platform — can’t learn that?

Then it’s not ready to work.
Not with me. Not with anyone.

email