A chat transcript proves something was said. When the thing talking back is a person, that's often enough. But when an agent revokes access, issues a credit, or reclassifies a customs entry, a transcript is no longer adequate. What matters is what was done, under whose authority, and how you challenge it.
Agents are already moving fastest in domains where work was review-shaped before AI arrived. Pull requests, incident investigations, support cases, compliance filings. These aren't chat-friendly because the institution already knows how to argue with them. A security operations team reviewing an agent's evidence bundle evaluates activity timelines, severity scores, and remediation guidance. The output is a structured investigation object, not a conversation.
So the unit of interaction becomes less like a message and more like an action with standing: mandate, scope, evidence, consequence, and a path for appeal. Call it a docket. The approval button evolves into a bound technical object showing what's about to happen, what supports it, and what breaks if it's wrong.
A docket-shaped review catches procedural and factual failures more naturally than emotional ones. The file knows who called and what was offered. It does not know whether the person felt heard.
Trace vs. record: A trace explains what the system did (model calls, tool calls, errors). A record explains why the act was allowed to stand (mandate, approval, scope, remedy).
The click problem: A human click once compressed presence, intent, authority, and timing into one gesture. Agents decompress it. The missing evidence has to live somewhere.
Approval decay: Too many low-quality approval prompts train people to stop reading. The result is autonomy by exhaustion, not by design.
Low-contest acts: Summarize, classify, draft, enrich. Easy to absorb.
Medium-contest acts: Route tickets, recommend decisions, open alerts. Update internal state.
High-contest acts: Spend money, revoke access, send customer-facing commitments, delete data. The organization must survive a challenge from a customer, auditor, regulator, or court.
What's unsettled: Protocols are standardizing how agents act (tool calls, task states). How institutions record what the act was and whether it should stand is far less developed.