Article

Why we built Squawk

By Keith Mangold · 2026-06-12

We are a small digital shop. We do not have a marketing department. What we have is a pile of content: blog posts, decks, recorded talks, RFCs, client memos, screenshots, half-finished ideas in Slack threads. Real material, most of it never seen by anyone outside the building.

For a long time the gap between "we make a lot" and "we post almost nothing" just sat there. We tried the obvious fixes. A scheduler moved posts onto a calendar, but we still had to write every post first, which was the actual problem. An agency wrote things that did not sound like us. Templates and a ChatGPT tab produced copy that anyone could smell was generated, because the model had no idea who it was supposed to sound like.

None of those solved the real problem, which was never scheduling. It was turning what we already had into posts that sounded like our actual people, consistently, without it becoming someone's second job.

So we built the thing we needed. Squawk takes the content you already produced, keeps a voice profile for each person and the company, and drafts posts that read like a human wrote them. You review in minutes instead of staring at a blank box. Then it publishes natively to each channel.

The test we held ourselves to: if a post sounds like AI, we failed. The whole reason most AI-drafted content reads like AI is that the model does not know who it is writing for, what the company does, or what good looks like. Squawk wires those three things in by default.

We now run all of LSA Digital's social through Squawk, including Squawk's own. The feed is the proof. If you are wondering whether it actually sounds like a person, you are already looking at the answer.