Writing a reply worth sending
A generated draft gets you most of the way. The rest is knowing what a reply that lands looks like, the habits that close the gap, and where a general-purpose generator stops and a real AI agent takes over.
What a good reply looks like
Same frustrated message, two very different replies. The first is fast and forgettable. The second acknowledges the frustration, owns the problem, and gives one clear next step.
The message
“This is the second time I've asked and still nothing. I need an answer today or I'm escalating this.”
A rushed reply
“Sorry for the inconvenience. We're looking into it and will get back to you soon.”
A reply worth sending
Hi [name], you're right to be annoyed — asking twice and hearing nothing back is on us, not you. Here's where it actually stands: [the real status]. I'll have a firm answer to you by [specific time today], and I'll come to you rather than the other way around. Sorry it took a second message to get here. [your name]
Six habits that make a reply land
A generated draft gets you about 80% of the way. These are the habits that close the gap, whoever you're writing to.
Acknowledge how they feel or what they asked in the first line before you explain anything. Someone who feels heard reads the rest differently.
If the answer is no, or not yet, say so plainly and early. A clear no beats a warm maybe that wastes everyone's week.
Be honest about what you can and can't do. Over-promising means under-delivering later, and that's what erodes trust.
End every reply knowing exactly what happens next and who does it. Vague replies generate more replies.
A text isn't an email. Drop the greeting and sign-off on chat; keep them where they're expected. The medium sets the register.
AI sounds confident even when it's wrong about your details. You're the editor on every reply, not the rubber stamp.
If most of your replies go to customers, the best ones are worth reusing. Turn them into help center articles customers can find on their own, or let an AI agent answer the repeat questions for you.
A generator vs a real AI agent
A general-purpose generator drafts a reply fast, but it can't see your product, answer on its own, or get the reply to the customer without you pasting it across. For customer replies specifically, here's where the two diverge.
| Capability | Generator | Selvo agent |
|---|---|---|
| Drafting a reply | ||
| Drafts a written reply to any message | ||
| Knows your product and policies | Only what you paste in | Trained on your help center |
| Applies your voice | Re-describe it every time | Set once |
| Cites a source the customer can verify | ||
| Getting it to the customer | ||
| Copy-paste each reply by hand | Every single one | Never |
| Answers customers 24/7 on its own | ||
| Lives inside your support inbox | ||
| Hands off to a human when unsure | ||
| Trust & cost | ||
| Guardrails against made-up answers | ||
| Audit log of every reply | ||
| Price | Free / $20-mo per person | $0.10 per resolved chat |
