🔍 Read the full analysis: Four Ways To Make AI Agents More Decision-Oriented 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
Thorsten Meyer’s October 27 article recommends four ways to keep AI agents from generating more human work than they save: count review and follow-up time, test ideas without committing to all of them, set stopping rules for recurring tasks, and limit how often a coordinating agent interrupts. The advice draws on Anthropic’s October 16 study of its own workforce, which found that employee-directed agents often waited for human input and could produce more material than staff had time to read.
In an October 27 article, Thorsten Meyer outlined four rules for making AI agents more decision-oriented: account for the human work their outputs create, test ideas without adopting every one, set stopping conditions for repeated tasks, and restrict how often coordinating agents interrupt. The advice responds to findings from an Anthropic study of its own workforce, which Meyer says found that agents spent much of their working-session time waiting for people.
Meyer’s central point is that generating an answer is only part of an agent’s cost. People may still need to read, verify and act on the output. Before starting a task, he recommends identifying who will review it, who will use it and whether the resulting work is worth doing. A fast-generated report, for example, can still consume substantial staff time if several people must check it.
His second rule is to use agents for low-cost exploration while keeping only ideas that prove useful. He warns that small experiments can quietly become ongoing projects, each adding updates and review work. His third rule is to define an end date and a “nothing new” stopping point before automating a recurring task, and to pause it if nobody reads the results.
For agents that coordinate other agents, Meyer recommends an attention budget: silence by default and no more than three daily decisions for a person to consider. Each should include a recommendation, supporting evidence and the consequence of taking no action. The article presents these as Meyer’s operating rules, not as results from a controlled test of the four methods.
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
Keeping Agent Output Within Human Capacity
The recommendations address a practical problem for teams adopting agents: more output does not automatically mean more useful work. If people must review every result, resolve every question and maintain every automated task, increasing the number of agents can shift effort rather than remove it.
That makes decisions about what not to read or continue part of agent management. A stopping rule can prevent a recurring report from consuming attention after it stops producing new information. A review budget can direct people toward consequential decisions instead of routine updates. Meyer’s framework is most relevant to workers and managers who are already receiving frequent agent-generated reports, drafts or requests for approval.
Products Worth Considering
Jackery Explorer 300 Portable Power Station, 292Wh
Anker SOLIX C200 DC Portable Power Station, 192Wh, 200W
Apple AirPods Pro 3 Wireless Earbuds with Active Noise Cancellation
As an affiliate, we earn on qualifying purchases.
Anthropic’s Workforce Study
Meyer cites research Anthropic published on October 16, 2026, based on how its own employees used agents. According to figures reported in the article, by August 2026 employee-directed agents logged six times as many hours as the employees directing them, while agents spent about 80 percent of session time waiting for human input. The article also says Anthropic estimated that reading all agent output would take non-technical employees about half their working time.
The same account reports that the share of sessions devoted to “meta-work,” mainly managing agents, 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. Meyer cautions that the figures describe one company: Anthropic’s employees adopted the tools by choice, and its work practices may not reflect those of other organizations.
The article also draws on Anthropic customer stories to illustrate the value of human review. Contractor Tony Severino reportedly caught an error in an agent’s blueprint calculation and now asks it to show its work. Such examples are vendor-published accounts, not independent evidence that the approach produces the same results across businesses.
“Every update looks reasonable on its own. Together they become a job.”
— Thorsten Meyer, describing the pattern in his article
How Far the Findings Apply
The figures Meyer cites come from Anthropic’s own employees, and the article notes that the company’s staff chose to adopt the tools. It is not clear from the supplied material how closely the study’s methods or workforce represent other employers, or whether the reported patterns hold across different agent systems and job types.
The article’s four rules are practical recommendations rather than findings shown to reduce workload in a comparative trial. The source material also does not provide the full details of the remaining examples or evidence measuring the rules’ effects over time. Readers should distinguish the reported workforce data from Meyer’s proposed practices and from vendor-published customer claims.
Testing the Rules in Daily Work
Meyer’s proposed next step is for teams to apply the rules before expanding agent use: identify who will handle each output, decide which experiments merit continued work, and specify when recurring tasks should stop. For coordinating agents, teams can set a daily limit on requests and require each request to state the evidence and the cost of inaction.
The article does not announce a new product, policy or follow-up study. Whether the approach reduces interruptions or review time remains to be measured by the organizations that adopt it. The immediate test is whether teams can spend less time managing agent output while still catching errors and acting on useful results.
Key Questions
What are the four rules for managing AI agents?
Meyer recommends counting the human work that follows an agent’s output, running small tests without committing to every idea, setting end conditions for recurring jobs, and giving coordinating agents a limited attention budget.
What did Anthropic’s study report about agent waiting time?
As reported by Meyer, agents spent about 80 percent of session time waiting for human input in Anthropic’s employee-directed sessions. The findings concern Anthropic’s workforce and should not be treated as a universal rate.
Does the article show that these four rules have been proven to work?
No. Meyer presents them as practical recommendations based on his work and the attention problem described in the article. The supplied material does not report a controlled evaluation of the four rules.
Why should teams set stopping rules for recurring agent tasks?
A recurring task can continue producing updates after it stops adding useful information. Meyer recommends setting an end date, stopping when there is nothing new, and pausing a task if people are not reading its results.
Source: ThorstenMeyerAI.com
