Anthropic and Slack's Jamie DeLange on human-AI agent teams

The blog post is the second entry in an Anthropic series on human-AI agent teams; the first covered Anthropic's own internal efforts. Slack was chosen for this installment because the company had been thinking about human-AI agent teams long before AI agents actually appeared, according to the post.

DeLange joined Slack in 2017 and led the company's search and machine learning efforts, with a mission described as turning conversations in the workspace into organizational knowledge. That mission, as Slack tells it, did not turn channels themselves into knowledge. Messages piled up unstructured, and people repeated the same questions.

Anthropic argues that understanding everything and pulling out knowledge for work is impossible for humans, but AI agents are now able to do it. The stakes hinge on whether teams follow the recommended practices, such as making public channels the default and asking agents why something happened rather than only what happened, rather than on the technology alone.