It's not going to settle down

Waiting for AI to stop changing is a strategy with no end date.

A few weeks back my husband and I worked out that our tenth wedding anniversary lands in January. Ten years. It feels like yesterday and a lifetime ago that we got married and jetted off to Sweden to chase my long held dream of an authentic Aussie pub in the heart of Stockholm.

I'm fairly sure the Covid years are responsible for the Jedi mind trick.

So I reached for my phone and opened the travel agent Custom GPT I'd built for our last overseas trip. Added to the parameters, handed it our wish list and let it get on with things. Three days later we had a seven day trip to Fiji booked and the dogsitter organised.

That Custom GPT now has a deadline on it.

OpenAI is retiring Custom GPTs as a format and I have a bunch of them! Migrate them to plugins by 11 December or they stop being available. I'll migrate mine, of course I will, but that's a pile of jobs that turned up on my list without me asking for it.

AI tools change constantly, the ones you build inside them can be retired or restructured without your input. What survives a change like that is the knowledge itself: a clear, documented picture of who your business serves, how it sounds and why customers choose you. Keep that in a document you control and every platform change becomes a copy and paste instead of a rebuild.

There is no settling down

There's a train of thought I think a lot of people are sitting with: I'll get properly into AI once it all settles down.

I understand it. Every month there's a new tool, a new model, a new thing you apparently should have been using since March. Waiting feels sensible, it feels like the responsible, don't-waste-the-money option. Except settling down implies a finish line and there isn't one.

The tools that were essential eighteen months ago have mostly been replaced, renamed or retired. The ones you're using today will likely go the same way eventually, the version of ChatGPT you learn this afternoon won't be the one you open next year.

That's not a reason to panic. It's a reason to stop building on rented ground.

What actually survives

Migrating my travel agent GPT isn't the hard part, because the thinking lives somewhere else. The instructions I wrote, the preferences, how we like a trip structured, what we'd never book, the budget conversation we have every single time. All of it sits in a document on my own drive. Moving it across to a plugin is a quick job, not a project!

The ones who'll feel this deadline hardest are the people whose only copy lived inside the tool. Everything they worked out through trial and error, every prompt they refined, every bit of context that made the thing genuinely useful, all of it sitting in a container with an expiry date on it.

That's the whole difference. Own the knowledge, rent the tool.

So I'm rebuilding mine this weekend

Which is exactly why I rewrote the final module of Repeat Business Foundation, because the one thing that will save you is having your knowledge base outside of your AI platform. That way you just need to run through your new 'employee' induction with it and point it to your AI brain.

Converting a Custom GPT to a plugin is pretty seamless, it takes all of the info stored in it and in a couple of minutes it's done! The time it will take is really dependent on how many you have and if you don't have this information captured elsewhere now is the time to save outside the platform.

The new module is your AI induction. Your business brain created, written down once, kept outside the tools, handed over whenever you start somewhere new. Who you serve, what you sound like, why people choose you over the place up the road, the things you'd otherwise explain from scratch every time you open a new chat.

It also covers the bit everyone skips. When to update it, what should trigger a review and where to back it up. A brain that's eighteen months out of date will confidently tell AI things about your business that stopped being true two summers ago. You won't find out until someone turns up expecting something you don't do any more.

Someday isn't a day of the week!

That's the honest answer to waiting for things to settle. You can't out-wait it and you don't have to chase it either. You build the thing that doesn't move, then you plug it into whatever is current. No rented ground.

Someday isn't a day of the week, but Monday the 19th is! That's when I'm running a free masterclass: Be the Business AI Recommends. The six things AI needs before it'll confidently put your name forward and what to fix first. Repeat Business Foundation doors open the same day, with the rewritten module inside.

Bring your questions about what just changed. There'll be a few!

Book your spot in the FREE Masterclass HERE