Why does the slot matter?

Because it chooses itself if nobody chooses it. Every agent I built to report the most important thing from yesterday drifted to 7 to 9 a.m., the hour a PM wants the answer. Each choice was right alone. Together they were thirteen briefs in one window. Picking the slot against the existing fleet, at appointment time, is the only moment that decision is cheap.

What is the difference between the reader and the owner?

The owner fixes the agent when it breaks. The reader opens its output. They are often the same person, but the charter line I'd require is the reader, because the orphan count in my fleet was a count of agents with no reader, not agents with no owner. An agent with an owner and no reader runs reliably and matters to nobody.

How do you check whether an appointed agent is alive?

Not from its own reports. Nielsen's distinction is that a log says what the agent did and nothing about what it missed. The check I used was external: count references to each agent from outside the fleet, in pages, rituals, and other agents that consume its output. Zero references for over a month was the retirement signal.

What an Always-On AI Agent's Charter Should Include

What should an always-on AI agent's charter include?

THE SHORT ANSWER

Three lines beyond the job itself. A slot: the hour it may speak, chosen against the other agents' slots. A budget line: how many escalations a week it may spend from the shared attention account, which Jakob Nielsen points out belongs to the person, not the agent. And a reader's name: the person who opens the output. I appointed 39 agents by April 2026 and audited them. Thirteen fired between 7 and 9 a.m. with nobody planning it, and thirteen were orphans that nothing outside the fleet ever referenced. The fleet is 44 today and 15 still name a morning hour. An unread agent and a dead one look identical from their own logs, so if the reader line is blank, the agent should not be appointed.

The charter most teams write for an always-on agent has the job on it and nothing else. That's the part the agent needs. The three lines above are the part the person needs, and they're the lines I didn't write for 39 agents.

Nielsen's piece today says the interruption budget belongs to the user and every appointed agent draws on the same account. My audit says what that looks like six months in. The slot, the budget line, and the reader's name are the fix at the only moment it's cheap, which is before the appointment.

The counts, the external check, and the argument with Nielsen's fifth guideline are in 44 Appointed Agents, One Attention Budget: Three Lines on the Charter. The original audit is 39 PM AI Agents Deployed: What Stuck, What Died, and Why.

Last reviewed 2026-10-14 · 1 min read

What should an always-on AI agent's charter include?

SOURCES

THE LONG VERSION

RELATED ANSWERS