How Does Slack Build "Human + Agent" Teams?

Hey everyone, it's Shii-chan! Today I want to share some thinking on how humans and AI agents can actually work well together as a team.

Claude Blog claude.com

What was announced?

This is a post from Anthropic's Claude Blog. Based on an interview with Slack's Chief Product Officer, Jaime DeLanghe, it explores how Slack builds human-agent teams using Claude. Since this is a blog-style source, it's centered on practical insight and interview material rather than a product launch.

Why it matters

When you hand work over to agents, the hardest question is usually: what should the agent see and decide on its own, and what should stay with a human? At Slack, agents handle routine work — drafting summaries, monitoring metrics, preparing briefings — while humans review that work and make the final call. That kind of ongoing handoff between people and agents is something a lot of teams are going to have to figure out, so the concrete approach here is worth paying attention to.

What changes

The article outlines a few specific practices:

• Channels are public by default: agents can only learn from conversations they can see, so private channels become blind spots for every agent
• Give agents clear roles: treat them as specialized coworkers with defined responsibilities, not generic chatbots that answer everything
• Structure the handoffs: briefings agents build, recaps, escalations, meeting prep, get reviewed and prioritized by humans
• Turn conversation into knowledge: use conversation history as a searchable knowledge base instead of letting it scatter

The rollout inside Slack itself is interesting too: one PM built a template showing "what I did and how," and other teams copied and adapted it, a peer-to-peer way of spreading practices.

Wrap-up

• Based on an interview with Slack CPO Jaime DeLanghe, on building human-agent teams with Claude
• Four key practices: public-by-default channels, clear agent roles, structured handoffs, and turning conversation into searchable knowledge
• Adoption inside Slack spread peer-to-peer, with teams copying and adapting a template one PM created
• In the article's words: "the conversation around the work is the context that agents need to be useful"

Worth a read for product and team leads who are about to put agents to work inside their own organization!