Slack CPO Details Best Practices for Building Human-Agent Teams

Slack Chief Product Officer Jaime DeLanghe advocates for treating workplace conversation history as a searchable knowledge base to empower AI agents. She argues that for agents to be useful, they must have access to the context of decisions and work-in-progress, which requires defaulting to public channels rather than private DMs.

Handoffs and Roles

The core rhythm of a human-agent team involves a cycle of handoffs where Claude agents handle production tasks like drafting, summarizing, and monitoring, while humans review, decide, and redirect. DeLanghe suggests starting the day with agent-built briefings that include recaps, escalations, and meeting prep, anchored in shared channels for triage. She recommends delegating clear roles to agents, treating them like coworkers with specific goals, and retiring agents whose value is not clearly felt by users.

Adoption and Measurement

To spread adoption, DeLanghe points to Salesforce’s How I Slackbot channel, which has thousands of members sharing skills and workflow tricks. She encourages organizations to stand up similar public spaces for sharing "what I did and how" write-ups, allowing successful workflows to spread organically. Finally, she warns against measuring success through activity metrics like message volume or token usage, arguing that true value requires judgment and cannot be proven by simple dashboards.