LinkedIn replies without losing your own voice

The challenge

Publishing on LinkedIn works — and fills the inbox: thank-yous, congratulations, expressions of interest and quick questions. Replying late lets warm leads cool down; replying to everything by hand eats the best hours of the day.

How it was done before

Messages were answered in batches at the end of the day, or not at all. In a hurry, the tone drifted: sometimes too dry, sometimes too long.

What we built

• Every incoming message is classified: thank-you, interest, simple question, or "Fermín needs to see this".
• Routine ones get a short, natural reply written in his voice — calibrated on real examples of his own replies, not templates.
• Anything that smells of negotiation, pricing or a complaint is set aside with no automatic draft.
• Every reply is logged so tone can be reviewed and corrected.

Human oversight

The system only answers the routine on its own. Any conversation with commercial intent or sensitivity goes straight to Fermín, and his corrections re-calibrate the assistant's voice.

What the client takes away

• Message classifier
• Own-voice replies with a review queue
• Full conversation log

Outcome

In daily use at Synkorai since 2026. We do not publish numbers we cannot show: if you want to see it working, we will show you live in the first meeting.

Limits

• It does not negotiate, promise terms or book meetings: that is Fermín's job.
• It only replies; it never starts conversations or prospects.
• The voice was calibrated on real examples and is reviewed periodically.