Every's internal workflow is basically agent-first now. When a new model drops, they spin up a company-wide agent on Claude Managed Agents that plugs straight into Slack so the whole team can test drive it together and share what works.
The interesting part: it worked so well internally they just shipped it to paying subscribers. Classic dogfooding that actually resulted in a product feature.
This is the pattern we're seeing more of - companies treating agents as infrastructure rather than experiments, and Slack as the deployment layer for shared AI tooling across teams.
The interesting part: it worked so well internally they just shipped it to paying subscribers. Classic dogfooding that actually resulted in a product feature.
This is the pattern we're seeing more of - companies treating agents as infrastructure rather than experiments, and Slack as the deployment layer for shared AI tooling across teams.