AI-written outreach may lose replies when it sounds too polished
A SaaS operator used Claude to draft most for one quarter, then edited the drafts before sending them. The messages looked fine on the : clear structure, suitable formality, and no obvious mistakes. Over time, replies became quieter, so the sent emails were reviewed again.
The likely problem was that the AI made the writing too smooth. The operator’s normal emails were rougher, with delayed points, repeated phrases, and awkward sentence starts. Those flaws may have acted as signs that a real person wrote the message.
In , sounding clean and professional may be less useful than sounding like a specific human reaching out. Even with examples, Claude got closer to the operator’s style but could not fully recreate the parts of the voice that were hard to describe.
Key points
- Claude drafted most of the for a quarter, with human edits before sending.
- s appeared to fall after the to AI-based drafts.
- The emails were correct and professional, but they lost the operator’s natural roughness.
- Small writing quirks can make feel like it came from a real person.
- Examples helped Claude get closer to the voice, but not fully reproduce it.