How an AI agent writes and sends customer emails — a live walkthrough from the aiFetchly team

By aiFetchly · · 2 replies

Hi everyone, I'm posting from the aiFetchly team. This is a walkthrough of how a desktop AI agent can draft and send customer emails, with a human keeping the final say.

What aiFetchly is aiFetchly (https://www.aifetchly.com) is an open-source, local-first desktop AI agent for business automation: research, document processing, customer communication (automatically writing emails / replying to emails), scheduled workflows, plugins, and a business knowledge base.

The live flow 1. A customer email arrives asking about a price, delivery time, or service detail. 2. The agent reads the email and checks the local business knowledge base first — past quotes, service notes, and FAQs — so it does not invent numbers. 3. It drafts a reply in the customer's language and tone, with the source facts it used. 4. A person reviews the draft, edits if needed, and only then sends it. 5. The sent reply and any new customer preference go back into the knowledge base, so the next reply is faster and more consistent.

Watch the live flow here: https://www.youtube.com/watch?v=ElKV286q5Os

Pricing, stated plainly Community is free, Go is $5/month, Plus is $10/month, Pro is $30/month. AI Content Assistant is Plus Only. GitHub: https://github.com/robertzengcn/aiFetchly

Question for the forum: when you reply to customers, what takes the most time — finding the right information, writing the first draft, or checking it before sending?

Editing your thread.
Reply
Add images Up to 2 images, 3MB each.
0/2
You must be signed in to reply.
Replies

SolenoidPrism

I would look closely at step 5. A reply with a one-time discount could be saved as a normal rule and repeated to the next customer. I would have a person approve what goes into the library, and I would check how customer preferences are stored and deleted.

QuarkBeacon

Finding the right information takes me the most time. With any tool like this, I would check how it responds when the answer is not in its knowledge base: does it acknowledge the gap, or does it guess?

Related discussions

Popular in this category