Agentic AI: findings that name their record, run text, History, copying and autonomy

A finding names and opens the record it is about; run summaries show real names again; running an agent is not an edit; History says publish and change-autonomy; Copy starts with a free code; Records changed (30d) and Produced (30d); a raise of autonomy stars its reason.

• What Agents Produced: a finding names the record it is about, from what the run read, and Open the record opens it on its own screen — for a reader who may open it.
• Run summaries, outputs and steps are stored with real values put back for you; the AI engine’s own conversation keeps the stand-ins.
• Running an agent no longer changes its Last modified. History records a Publish as publish (with the version and note) and a raise or lowering as change-autonomy (from, to and the reason).
• Copy starts with a code that is free — the original’s with -COPY, then -COPY-2 — and “(copy)” on the name.
• Agents shows Records changed (30d) — records created or changed in other modules — beside Produced (30d) — findings, proposals and messages.
• Change autonomy stars Reason as soon as a higher level is chosen; a reason of spaces alone is refused, and the reason is kept trimmed.
• An agent whose saved model is no longer offered is marked on the list (Model problem); Run it and Publish say “Choose a valid model first”.
• Model and Owner on the Agent form are starred with what they are needed for — “needed to publish or run” — though the agent still saves without them.
• An agent’s 360° cards open only that agent’s runs and workflows, and their items say what they are.

Since 19 October: Switch off and Switch on, cost ceilings in the company currency, Tool calls and Calls succeeded, and a Code fixed once created. See Agentic AI: Switch on and off, cost ceilings, the model it will use, tool calls, a fixed code and a local name.