This analysis is based on observations from last week about the GPT-5 launch and its impact on business users. The aim is to better understand how to navigate the AI transition without getting lost in the chaos.
Update:
Updates to ChatGPT:
You can now choose between “Auto”, “Fast”, and “Thinking” for GPT-5. Most users will want Auto, but the additional control will be useful for some people.
Rate limits are now 3,000 messages/week with GPT-5 Thinking, and then extra capacity on GPT-5 Thinking…
— Sam Altman (@sama) August 13, 2025
Last week, OpenAI launched GPT-5 in a way that caught everyone off guard and not in a good way. They removed access to all previous models (GPT-4, GPT-4o, o3) without any warning. Users simply woke up to find the version of ChatGPT they’d been using daily had vanished.
What actually happened
OpenAI approached the launch like a research company, not a consumer products company. They essentially said: “We’ve got something new and technically better, so we’re replacing everything old.” No transition period, no option to delay the update, no clear documentation about changes.
Compare this to how Apple launches a new iOS: weeks of notifications, beta versions for testing, the ability to stay on the old version if you want. It’s a fundamental difference in approach.
Why This Matters for Your Business
This launch shows us something important: in the AI world, technical superiority isn’t enough. OpenAI probably has the best AI researchers, but that doesn’t automatically make them good at delivering stable products for business.
For your business, this means being prepared for:
- Abrupt changes in the tools you use
- Lack of predictability in launches and updates
- The need to have alternatives ready
How to Approach This Practically
Use different tools for different tasks Personally, I use Claude for writing (sounds more natural), GPT-5 for coding and complex analysis, and sometimes local models for sensitive stuff. Don’t put all your eggs in one basket.
Build processes that can survive changes Instead of creating super-specific automations for one particular model, make them flexible enough that you can switch providers if needed.
Document decisions, not just implementations Write down WHY you chose a certain approach, not just HOW it works. That way, when the tool changes, you know what needs to be preserved.
The Good News About GPT-5
To be fair GPT-5 does bring serious improvements:
- Costs less per token processed
- It’s faster
- Makes far fewer mistakes (the famous AI “hallucinations”)
The problem wasn’t the product itself, but the completely unprofessional way it was launched.
What You Do From Here
The AI landscape will continue to change rapidly. That’s not necessarily bad it creates opportunities for continuous optimisation and improvement. You just need to be prepared.
Focus on:
- The results you need, not the latest model launched
- Stability in critical processes
- Flexibility for experimentation in less critical areas
- Maintaining perspective on which changes actually matter
You don’t need to adopt every new model the moment it appears. Your value (or your consultant’s) lies in filtering the noise and adopting the right technologies, at the right time, in the right way.