Every LinkedIn comment answered in the founder's voice, with a human on send

Problem

The founder's LinkedIn posts pull real comment volume, but replying to each one in his own voice took time he did not have. Any tool that logs into his account to read or post comments also risked getting it flagged or banned.

What I built

• Comment capture with Apify, cookieless, so nothing ever logs into the founder's LinkedIn account and the posting risk stays off it.
• A style guide built from the founder's own comment history and tone, feeding a language model that drafts a reply to every new comment in his voice.
• Drafts land in an Airtable approval queue, one row per comment, where a team member reads it, approves or edits the wording, and posts it by hand from the founder's account. Nothing goes out without a person reading it first.

Result

In testing, a post with 283 comments including replies was captured in full, 166 of 166 threads, for 36 cents. The queue now handles 50 to 1,000-plus comments a week at a running cost of about 5 to 25 dollars a month. The founder's replies keep going out at volume without him writing each one or putting his account at risk to get them.

What it shows

A language model can carry the drafting load on something as personal as a founder's voice, as long as a person still reads and sends every reply, and the scraping step is built cookieless from the start instead of bolted on as an afterthought.