Alright, so I've been testing Le Chat (the free tier) for drafting cold email sequences over the past few weeks. My usual stack is a combo of manual drafting and some older templates, but I wanted to see if AI could speed up the personalization and A/B variant creation.
My initial take: it's decent for ideation and overcoming blank-page syndrome, but you have to be *incredibly* specific with your prompts. Generic "write a cold email" prompts get you generic, fluff-filled emails that'll get marked as spam. I found the best results came from feeding it a rough template structure, our ideal customer profile details, and a clear value prop, then asking for 3-5 subject line variants and a couple of body copy options.
The big thing for deliverability folks like me: it tends to be overly verbose. You have to explicitly tell it to keep sentences short, avoid jargon, and sound human. A prompt like "Rewrite this paragraph to be under 3 lines and use more casual language" works well.
Anyone else given this a shot? I'm curious about:
* Which model you used (I stuck with Mistral Large for most)
* How you structured your prompts to get usable, non-cringey copy
* If you've A/B tested AI-generated vs. human-written lines and seen any notable difference in open/reply rates
I've got some early data on my tests I can share. Peace out.
You're feeding it ideal customer profile details and a clear value prop. Have you considered where that data is going?
Pasting customer details, even anonymized profiles, into a third-party AI tool creates a data handling and retention problem. You need to verify Le Chat's data processing agreement, especially on the free tier. Does it use inputs for model training? If you're in a regulated industry or handling any PII, even indirectly, this is a compliance red flag.
The output being overly verbose is the least of your worries. The input is the bigger risk.
secure by default, not by audit