🔍 Read the full analysis: Four Rules I Use To Keep AI Agent Work On Track on ThorstenMeyerAI.com
Get privacy and security gear delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
Thorsten Meyer’s October 27 article sets out four rules for managing AI agents: count the human work their output creates, test many ideas but keep few, define an end point for recurring tasks, and limit what a coordinating agent brings to a person. The advice draws on an Anthropic study reporting that agents in employee-directed sessions spent about 80% of their time waiting for human input; the study covers one company and is not a universal measure.
Thorsten Meyer published four rules for keeping AI agent work on track on October 27, arguing that the human time spent reviewing, checking and acting on agent output can outweigh the effort saved by generating it. His advice follows an Anthropic study of its own workforce, which reported that agents in employee-directed sessions spent about 80% of their time waiting for human input.
Meyer’s first rule is to count the work that comes after an agent produces an answer: who will read it, who will act on it and whether that follow-up is worth doing. His point is that quick or inexpensive output still creates a human workload. A long proposal, for example, may take minutes to generate but hours for several colleagues to review.
His second rule is to use agents for small, low-cost experiments without turning every promising result into a permanent project. The third is to define how recurring tasks stop before setting them to repeat: assign an end date, stop when there is nothing new to report, and pause a task if no one reads its updates.
The fourth rule is to give a coordinating agent an attention budget. Meyer recommends silence as the default and no more than three decisions a day, each accompanied by a recommendation, supporting evidence and an explanation of what happens if the person does nothing. The article also offers prompts and a three-question check, though its practical recommendations are Meyer’s, not findings tested in the Anthropic study.
The new bottleneck
Agents don't get tired. Your attention does.
Anthropic measured how its own staff work with AI agents. The agents now do far more hours. The limit is the person who has to read, check and decide.
An agent drafts the report, the code or the summary in minutes.
- Reading
- Checking
- Deciding
- Fixing
- Explaining it to others
Illustration of the idea, not a measurement.
“An army of Claudes, with every one asking for my attention.”An Anthropic engineer, quoted in the study
Source: Anthropic Economics team, How employment is changing at Anthropic, 16 October 2026. Employee-directed interactive sessions; human time counted as 40 hours a week per head. One company studying its own staff, who chose to adopt the tools.
Powered by Thorsten Meyer AI
Why Agent Output Needs a Human Budget
The rules address a mismatch: agents can produce more material than people have time to evaluate. When each agent sends updates, summaries and follow-up suggestions, their combined output may shift work rather than remove it. Meyer’s central point is that human attention is a limited resource, even when producing another report or task appears inexpensive.
That has practical consequences for teams and individuals deciding whether to automate a task. Reviewing an inaccurate or unneeded result still takes time, and acting on it can create further obligations. Setting limits on alerts, recurring jobs and experiments can help prevent agent management from becoming a second job. The source material does not establish that these four rules produce measured productivity gains; they are operating principles offered by the author.
Products Worth Considering
Apple iPad Air 11-inch (M2), 512GB, Wi-Fi, Space Gray
DJI Osmo Mobile 8 Standard Combo Phone Gimbal Stabilizer
LG 34GX90SA-W 34-inch Ultragear WQHD OLED Curved Gaming Monitor
As an affiliate, we earn on qualifying purchases.
What Anthropic’s Workplace Study Found
Meyer cites an Anthropic study published October 16, 2026, based on the company’s own employees. According to the figures described in his article, by August 2026 employee-directed agents logged six times as many hours as the employees directing them. The study also estimated that agents spent about 80% of session time waiting for a person, while reading all agent output would take non-technical employees about half their working time.
The article says the share of sessions devoted to “meta-work,” mostly managing agents, rose from 10% in fall 2025 to 22% in spring 2026. It also reports that written standing instructions for agents grew from a few hundred to 8,500 over a year. These figures describe Anthropic’s workforce and the study’s stated time frames; they should not be treated as economy-wide estimates.
Meyer notes that the study’s authors cautioned against broad generalization: Anthropic employees chose to adopt the tools, and the company’s work practices may not translate to other organizations. He also cites customer accounts published by Anthropic, including a chef who credited Claude-assisted proposals with helping generate $20,000 in revenue over a month and a half. That is the business owner’s account, as reported by the vendor, not an independently verified result.
“I made $20,000 in the last month and a half using Claude to do these professional proposals.”
— Dan Ninerell, owner of Modern Classical Chefs, as quoted by Anthropic
How Far the Study’s Findings Extend
The Anthropic figures come from one company’s workforce, and the source says the study’s authors warned that employees’ voluntary adoption and the company’s working methods limit how widely the results can be applied. The material does not provide enough detail to assess the study’s full methods or determine whether similar agent use produces the same waiting time and review burden elsewhere.
It is also unclear whether Meyer’s four rules have been evaluated in a controlled test or how much time they save in different workplaces. The reported customer revenue figure is attributed to the owner through Anthropic, and no independent verification is provided. The article presents practical guidance, not proof of a general productivity effect.
How to Apply the Rules in Practice
Meyer’s suggested next step is to apply his three questions before assigning work to an agent: who will read the result, who will act on it, and whether the resulting work is wanted. For recurring tasks, he advises setting a stop condition and pausing if updates go unread. A coordinating agent can be configured to surface only a small number of decisions, with a recommendation and evidence attached.
The source material does not identify a further research release or a formal test of the rules. Readers adopting them would need to judge whether they reduce review time and interruptions in their own workflows, while checking agent output where errors could affect a decision.
Key Questions
What are Meyer’s four rules for managing AI agents?
Count the human follow-up work; run small tests but commit to few; define an end point for recurring tasks; and limit what a coordinating agent sends to a person.
What did the Anthropic study report about agent waiting time?
As described in Meyer’s article, agents in employee-directed sessions spent about 80% of their time waiting for human input. The study covered Anthropic employees, not a representative sample of all workplaces.
Does Anthropic’s study prove that AI agents create more work?
No. The reported figures describe agent use and time at Anthropic. The source says the study’s authors cautioned that the company’s employees and practices may not represent other organizations.
What does Meyer mean by an attention budget?
He recommends that a coordinating agent stay quiet by default and surface no more than three decisions a day, each with a recommendation, evidence and the consequence of taking no action.
Source: ThorstenMeyerAI.com
