Generative AI is a useful tool, but I enjoy writing to my clients. I think they appreciate the personal replies too. While I might use it to automate mundane tasks, I prefer the human touch both when sending and receiving correspondence.

Each to their own, of course.

You are right that it needs to be double-checked. I recently, for fun, got Copilot to write a document, then got it fact-checked by ChatGPT, which found errors. It corrected the errors, and I ran it through Google. It found more faults. I fed that back into Copilot and asked it to check, and it found fault with the corrections.

Thanks for taking the time to comment.

The confidence problem can be worse than the accuracy problem. When Copilot gives the wrong operating temperature for your camera, it delivers that answer in exactly the same tone and format as when it’s correct, with no signal in the output to tell you which situation you’re in. The cross-checking approach described here is the sensible response, but it eats most of the time-saving the tool was supposed to deliver.

I use AI for client correspondence and the admin side of running a photography business. For broad questions it holds up well enough. For anything specific (particular gear specs, venues I regularly work with through LemonLens, current supplier pricing) checking is still necessary. A tool that confidently gives the wrong answer to a specific question can be more hazardous than no answer at all.