Companies are building increasingly capable AI employees (agents) - and we're struggling to manage this growing digital workforce.
@takidau CTO of
@redpandadata and author of the O'Reilly Streaming Systems book, came on
@chain_ofthought to explain why he thinks agent deployments stall out in the prototype-to-production gap. His take: it's not a model problem, it's a governance problem - nobody has built effective HR for agents.
Inline guardrails like your CLAUDE.md work until they don't. Tyler's analogy: it's like telling a human all the company cash sits in a room with no lock and asking them not to take any. Real governance has to live out-of-band, in channels the agent can't see, modify, or override.
He walks through the four pillars (identity, authorization, observability, accountability) and a wealth-management demo where any trade over $1,000 gets routed to a human and the agent doesn't even know the rule exists.
Watch the full episode 👇
Chapters:
0:00 Why nobody built HR for AI agents
2:12 Three ways agents differ from human employees
7:53 The four pillars of out-of-band governance
10:29 Identity: task-scoped, short-lived
14:40 Authorization: deny-capable and intersection-aware
18:57 Observability: record everything via OpenTelemetry
24:24 Redpanda's agents and the $1,000 trade limit example
30:10 Accountability and the kill switch
34:02 The Agentic Data Plane
41:20 Should we stop chasing model alignment?
44:04 Building human-like value systems into agents
47:25 Tyler's 12-24 month outlook