AI is generating more work output than ever, which creates more "workslop" (polished looking artifacts that shift the real effort onto whoever receives them). Most organizations deploy AI as private tools for individuals, so its output arrives with no context, no real accountability, and no real way to build on it -- actually having the opposite effect of creating more work for the folks receiving AI generated output.
Drawing on how we built and rolled out agents in our organization with exactly that theme in mind: AI teammates that work in the open in our Slack, learn from correction, and operate within the permissions people already have. I'll share what changes when agents are accountable the way teammates are, why that structurally prevents workslop, and a practical path for engineering leaders to move their orgs from tools to teammates
