🔍 Read the full analysis: Stop Babysitting Your AI Agents: The 4 Rules I Use To Get Decisions, Not Updates on ThorstenMeyerAI.com
Get the latest gadgets 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
In an October 27 essay, Thorsten Meyer argues that AI agents can create more human work through updates, reviews and follow-up. He draws on Anthropic’s study of its own workforce and proposes four practices: account for downstream work, treat experiments as tests, set stop conditions for recurring tasks, and limit how often a coordinating agent asks for attention.
Thorsten Meyer published a set of four rules for managing AI agents on October 27, arguing that workers should focus less on generating agent output and more on limiting the human time needed to review and act on it. His essay cites an Anthropic study of its own workforce, which reported that agents in employee-directed sessions spent about 80 percent of their time waiting for human input.
Meyer’s central point is that quickly generated work still carries a cost: someone must read it, check it and decide what to do. He recommends asking who will handle those steps before launching an agent. If the output creates work nobody wants or has time to do, he argues, the agent may not be saving time overall.
His other rules are to use agents for small experiments without turning every promising result into a permanent project; define an end date and “nothing new” stopping condition for recurring tasks; and set an attention budget for a coordinating agent. Meyer suggests that such an agent default to silence and surface no more than three decisions a day, each with a recommendation, supporting evidence and the consequence of taking no action.
The article also describes examples from Anthropic customer stories. South Jersey chef Dan Ninerell said he used Claude to draft catering proposals and attributed $20,000 in business over a month and a half to that work. Chicago-area contractor Tony Severino described checking Claude’s blueprint calculations after catching an error in an early bid. These are individual accounts published by the vendor, not independent evaluations of the tools.
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 recommendations address a practical problem for people adopting agent tools: the software can make producing drafts, reports and alerts easier, while leaving the review and decision workload with people. If that follow-up is not planned, more automated output can mean more interruptions rather than less work.
Meyer’s proposed limits are operational rather than technical. Teams can decide what counts as a useful result, how often they want updates, who verifies important claims and when a recurring workflow should stop. Those choices matter because an agent’s answer may look complete while still requiring human scrutiny, as Severino’s blueprint example illustrates.
The essay’s prescriptions are Meyer’s own practices, not findings tested in a controlled study. Their value to other teams will depend on the kind of work involved, the risk of an error and the time required to check results.
Products Worth Considering

Apple MacBook Air 15-inch (2026) with M5 chip, 16GB memory, 512GB SSD — Midnight

Bose SoundTouch 30 Wireless Speaker, Black

Logitech MX Brio Ultra HD 4K Webcam
As an affiliate, we earn on qualifying purchases.
The Anthropic Figures Behind the Advice
Anthropic’s research, published October 16, examined how its employees used AI agents. As summarized by Meyer, by August 2026 employee-directed agents logged six times as many hours as the employees directing them. The study also found that agents spent about 80 percent of session time waiting for a person. Meyer reports that Anthropic estimated reading all agent output would take non-technical employees about half their working time.
The study also tracked work involved in managing agents. Meyer says the share of sessions spent on “meta-work,” mostly agent management, rose from 10 percent in fall 2025 to 22 percent in spring 2026. Written standing instructions grew from a few hundred to 8,500 over a year, according to the essay’s account of the study.
Those figures describe Anthropic’s employees and workplace, not all companies or agent users. Meyer notes that the study’s authors cautioned that employees adopted the tools by choice and that Anthropic’s working methods may not transfer broadly. The data offer a case study, not a universal measure of agent productivity.
“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, in an Anthropic customer story cited by Meyer
How Broadly the Findings Apply
The Anthropic numbers come from one company’s workforce and do not establish how much time agents spend waiting or how much output workers must review elsewhere. The article does not provide enough detail to independently assess the study’s methods or compare its estimates with other workplaces.
Meyer’s four rules are recommendations, not a measured intervention. It remains unclear whether limiting a coordinating agent to three decisions a day, or applying the other practices, would reduce workload across different jobs. The business figures in the customer examples are also attributed claims, not independently verified results.
How Teams Can Test the Rules
Meyer’s advice is to apply the rules within existing workflows: identify who owns review and follow-up, set a defined end to recurring tasks, and silence agents unless they have a decision to raise. Teams adopting that approach would need to monitor whether it reduces unnecessary updates without delaying important information.
The essay does not announce a product, formal trial or scheduled follow-up. The next evidence would come from teams tracking the time spent reviewing agent output, the number of useful decisions surfaced and whether stop conditions prevent abandoned experiments from becoming ongoing work.
Key Questions
What are Meyer’s four rules for AI agents?
He advises teams to count the human work after an agent’s output, run experiments without committing to every result, set stopping conditions for recurring tasks and give coordinating agents a limited attention budget.
What did Anthropic report about agent waiting time?
According to the figures summarized by Meyer, agents in employee-directed sessions spent about 80 percent of session time waiting for human input. The finding concerns Anthropic’s own workforce.
Does the Anthropic study prove agents increase workload?
No. The reported figures indicate substantial waiting and management activity at Anthropic, but they do not by themselves prove that agents increase workload across organizations. The study’s setting and methods limit how broadly the results can be applied.
Are the customer revenue claims independently verified?
The $20,000 figure is Dan Ninerell’s account in an Anthropic customer story cited by Meyer. The article does not provide independent verification of that amount or show that the result is typical.
Source: ThorstenMeyerAI.com
