Why AI-powered pulse surveys catch burnout before it spreads.

A pulse survey is a short, recurring check-in. Two lines. Three at the most. A single number, or a single sentence, taken at a fixed cadence — in Moodkin's case, every Tuesday morning, dropped into a Slack channel or a Microsoft Teams DM, answered in under thirty seconds.

The format is intentional. Engagement scores from the major surveys — Gallup, the Engagement Institute, the annual mid-market census — are useful in a 1,500-person organisation with a dedicated people team. They are next to useless in a thirty-five-person team that needs signal this morning. The pulse asks one question, gets one answer, and stores the trend.

A quarterly survey falls apart for two specific reasons in a small team. First, the cadence is too slow. If burnout surfaces in week eleven, a survey administered in week twelve and analysed in week seventeen is telling you about a problem that has already cost you a great engineer, a great designer, or a great account lead. The signal arrives after the resignation.

Second, the granularity is too coarse. A 7.2 average across a forty-person company tells you almost nothing about the four people who have stopped enjoying it most. The same 7.2 can hide a sub-team whose pulse has slid from 7.4 to 5.1 across eight weeks while the rest of the company is steady at 7.3.

A weekly pulse inverses both problems. The cadence is fast enough that a three-week drift is visible while you can still act on it. The granularity is sub-team level by default, so a quiet pattern in one corner of the company surfaces weeks before the company-wide average moves.

A weekly number on its own is not enough. The strongest pulse systems read the open-ended responses too — the half-sentence free-text answers that arrive in the same channel — and classify them against three signals the product was built to catch.

• *
  Sentiment drift.When vocabulary like 'tired', 'stuck', or 'spinning in circles' starts to climb relative to words like 'shipped', 'clean', and 'ahead', the model flags it as a pre-burnout trajectory rather than a frustrating sprint spike.
• *
  Imminent flight risk.Tenure, manager tenure, sentiment trend, and unprompted 'considering' or 'LinkedIn update' vocabulary combine into a single team-level risk score. The output is never 'this person is at risk', only 'this sub-team's risk score is up meaningfully this quarter'.
• *
  Workload asymmetry.A 25% subset of the team carrying a 60% share of on-call, release, or incident load produces a vocabulary pattern the classifier picks up. No HRIS pause-and-resume button will surface this.

All three signals are explained back to the founder or people lead in plain language, with a single concrete ask attached.

A signal is only useful if a person acts on it. This is where most pulse systems stall — they hand a founder a dashboard and a chart and call it done. The follow-through is where products like Moodkin differ: they post a single concrete ask to the founder's DM every weekday morning, with a draft talking point for the 1:1 that should follow.

Three examples from real patterns we have seen this quarter:

• *
  Sentiment drift in a quiet sub-team.Rebalance on-call. Publicly thank the overloaded sub-team this morning. Schedule two unstructured coffees this week — one each with the quietest people, no agenda.
• *
  Imminent flight risk in one sub-team.Pull the last 1:1 notes. Coach the manager on a stay-conversation script. The model does not name names — the manager, with the talking points in hand, decides whether to make the call.
• *
  Workload asymmetry around a release.Move two people off the on-call rota. Push the release by a week if you can. Buy two coffees for the on-call handoff.

Each ask is sized to take fifteen minutes, not the whole afternoon. The point is not to fix the team — it is to keep the founder in the loop while a small action is still cheap.

The job of a founder or people lead is to listen. Most of us only listen when the signal is loud. Pulse surveys, with the right classification on top, let you listen on Monday — when the signal is still quiet enough that one good conversation resolves it.

Moodkin is built exactly for the under-200 team. If you have been catching burnout at the resignation Slack-message rather than the Monday before, the install takes an afternoon. We'll send the playbook and book a thirty-minute kickoff.