ChatGPT Work, the most confused product marketing ever?

Having used ChatGPT almost daily since it’s public unveiling as GPT 3.5 back in 2022, I’ve gotten used to the constant UI tweaks, slew of feature updates, and the odd u-turn on functionality. But the updates that came along with GPT 5.6 and the new ChatGPT app have left many scratching their heads.

I thought it was just me who just couldn’t quite unravel the thinking behind some of the changes, but a quick look across social media certainly shows that a lot of folks are similarly baffled. As a marketer I know all too well that the rapid release cycles in the world AI models makes the clarity of messaging difficult to keep pace. If anything though, it’s even more important to get right.

Typically, much of the video content and written material for 5.6’s launch was fairly slick as you would expect, but the cut-through messaging about the features in the new ChatGPT app is as confused as the design choices. Here’s just a few examples:

Codex & Atlas apps gone: They were decent apps and easy to explain to consumers. Now GPT is a bizarre mix of one app, but the normal chat experience (which most people only use) is still browser-based, and though Atlas technically lives on, you will likely never use it as has been relegated to a little window.

What is Work for?: On the back of Claude CoWork (which was an equally hard concept to explain to most users at first), GPT has done the same thing, except it does not even change the interface when switching between Work and Codex… bizarre!

Where do you actually chat?: Answers on a postcard, I’m still using the deprecated original Mac app as it was the nicest experience!

Unwanted Pets: Does anyone know what these are for? OpenAI can’t even acknowledge them let alone explain them. They were almost embarrassed to talk about them.

Five things every AI product marketing team needs to get right:

1. Natural next steps matter. Combat the velocity mismatch (where releases outpace marketing), and build an early narrative that clearly steers users on a tangible roadmap that they can understand and be comfortable with.
2. Consistent narrative. It’s ok to have live demos, social campaigns, videos and blog all go out at the same time… but not if the message doesn’t hammer home consistently. This is particularly acute when it comes to software features.
3. Good engineering isn’t always good product. When engineering and developer teams can do something, it does not always mean they should. Push boundaries yes, but never forget that data from customers should always inform your decisions to tweak.
4. Tangible not obscure. The flashy features or customer stories you think hit home with a broad base of users might not be the one you think it is. If you are making tools for broad adoption don’t hyper focus on the cleverest application as your hero story, go for commonality.
5. Slow down to speed up. Failing is positive, it’s how some of the world’s greatest inventions have come about. But in the AI software world we’re seeing a vast graveyard of wrong turns that simply didn’t need to happen. Understand what users really want to do, how they can be helped, and why your next release needs to feel intuitive and not like a maze.