https://harishgarg.com/writing/how-to-fine-tune-gpt-3-api/

There's very likely a SAAS product that does this this the way you want off the shelf. I'd look for that before considering implementing anything yourself.

I think what you’re looking for is a chat bot. Tons of off the shelf solutions and tutorials if you want to DIY

I would definitely save the data. The best chatbots are a mix of handcrafted rules and fine tuned classifiers.

Definitely possible. If you’re sending out the same thing over and over you would probably be better off just automating it without AI though

Before you go fine tuning LLMs on your conversations I would check any privacy policy in the communications you are using. Using the other party’s messages for an LLM that will be exposed to other people opens you up to some risk.

I do contract R&D, I can help you build this. DM me and we can talk details.

Never mind automating with ML, have you tried just saving templates that you can manually copy/paste?

Was going to say this. I have some experience in this area from industry and theres still a lot of handcrafted business rules that go into chatbot development.

This is also where I'd start.

You can set up outlook rules to sort inbound requests into different folders or tag them by keyword, too. Id still recommend doing that final step of copying in the appropriate response manually though.

Thanks for the reply, and great to hear this affirmed by you two. I put a lot of thought into my responses, and my gut tells me that throwing out that data would be a waste.

For example, my current messaging platform does have automated responses and very basic AI-generated responses... both of which are almost never actually useful in a real-world conversation, in spite of putting in a number of hours to set up those 'automated' responses. I believe that is, in part, due to the fact that the AI is not 'learning' from my existing and new conversations.