ChatGPT, Copilot or Claude: which one for which job

Teams rarely choose one assistant. They end up with Copilot because it came with the Microsoft licences, ChatGPT because someone expensed it, and Claude because a developer preferred it. That is workable, but only with a rule about which is for what. Without one, people default to whichever they opened first and get worse results than the tools are capable of.

Copilot, for work inside your own files

The reason to use Copilot is not that it writes better. It is that it can see your tenant: your documents, your mail, your meetings, subject to the permissions you already have. For anything that starts with "based on last quarter’s report" or "summarise this meeting", that access is the whole advantage.

ChatGPT and Claude, for work you paste in

When the material comes to the assistant rather than the assistant reaching into your systems, the choice is about how the tool handles the task. Longer documents and careful reasoning over a large pasted context tend to go well in Claude. Fast iteration, image work, and a broader plugin ecosystem tend to go well in ChatGPT. Both are strong writers; the difference is smaller than the marketing suggests.

The rule we teach

• ✳If the answer depends on internal files, use the tool that can see them
• ✳If you are pasting the material anyway, use whichever your team writes better with
• ✳Never paste confidential material into a tool the company has not approved
• ✳Write the rule down; a rule that lives in one person’s head is not a rule