I’d bet a good chunk of agent spend pays for work that was wrong the moment it started.

@serroai goes after that problem. They just launched an “Agentic TPM (Technical Program Manager)” that updates coding agents when the decisions behind their work change.

Serro treats coding-agent failure as a coordination problem, since agents keep executing decisions that people have already changed in meetings, chat or tickets.

To catch those changes, it connects Slack, Linear, GitHub, Jira, Zoom, Notion, email and the codebase, then maintains a versioned record of each program’s decisions, owners, risks and history.