On 7 August 2025, Microsoft announced that GPT-5 is available in Microsoft 365 Copilot, on the same day OpenAI launched the model. I have now used it daily for almost two weeks, and I want to share what I have noticed.
1️⃣ Better at the hard questions
The biggest difference is on complex requests: long documents, multi-step questions, or prompts where Copilot needs to combine several sources. Answers are more structured and more complete, and I need fewer follow-up prompts.
2️⃣ Quick when it should be quick
Not every question needs deep reasoning. Copilot decides when a fast answer is enough and when it should think longer. As a user I do not have to choose a model; I just ask.
3️⃣ Better writing and better summaries
Emails, meeting summaries and document drafts feel more natural and need less editing. For people who use Copilot mainly for writing, this is where they will notice GPT-5 first.
What it means for your organisation
- No new project needed. The model arrives inside the Copilot your users already have.
- Same security and compliance. Copilot still works within your Microsoft 365 permissions and data protection.
- Retrain your champions. Better models open up use cases that did not work well before, so revisit the prompts and scenarios in your adoption programme.
My tips
- Try again the complex prompts that disappointed you before.
- Give context: who the output is for and what format you need.
- Keep reviewing the output. A better model is still not a replacement for your judgement.
My take
The speed at which Microsoft brings new models into Copilot is impressive, and it is one of the strongest arguments for Copilot: organisations get the latest AI inside the tools and security boundary they already use.
Have you noticed a difference since GPT-5 arrived in Copilot? Tell me what improved for you. 🚀
Leave a Reply