Automated Slack Metrics and Anomalies
Operations at Whop on using Claude to ship product & automate ops
This turns the daily metrics post from a passive dashboard mirror into a lightweight management layer. At Whop, Cowork is not just copying numbers into Slack. It is adding a first pass on what changed, what looks abnormal, and what deserves attention, so teams can start with an interpretation instead of a raw chart. The summary is still only the entry point, because verification and debate happen in the Slack thread and in the underlying dashboards.
-
Whop uses recurring Slack metric threads where Cowork posts the numbers every day, plus a sentence on what matters. When Cowork flags an anomaly or spike, the team does not treat that as final truth. The post kicks off discussion, and people who watch those metrics daily pressure test the interpretation.
-
This matches a broader pattern in operational AI. Internal Slack updates are where teams are most willing to let an agent run automatically, because the output is easy to challenge, delete, or refine. External communication, money movement, and compliance still keep a human reviewer in the loop.
-
The real product value is not anomaly detection by itself, it is compression of analyst work. Instead of someone opening dashboards, writing a recap, and telling the team what to inspect, the agent does that first pass automatically. Humans then spend time on root cause and action, not formatting the report.
The next step is a fuller operating loop where the agent not only posts the spike, but also links the likely cause, cites the source dashboard, and drafts the next check to run. The companies that win with this pattern will treat the agent as an always on junior analyst, with humans owning final judgment.