🔍 Read the full analysis: What I Learned From Asking AI Agents For 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
An October 27 article argues that AI agents can shift work onto people by generating more output than they can review and act on. It cites Anthropic’s study of its own workforce and recommends budgeting human attention, limiting recurring tasks and asking agents to surface decisions rather than routine updates.
In an October 27 article, Thorsten Meyer argues that people using AI agents should ask for fewer routine updates and more decision-ready recommendations, citing an Anthropic study of its own workforce that found agents spent much of their session time waiting for human input. The analysis focuses on a growing practical concern: agents may produce work quickly while leaving people to review, verify and act on it.
Meyer reports that by August 2026, employee-directed agents at Anthropic logged six times as many hours as the employees directing them. In those sessions, agents spent about 80 percent of the time waiting for a person, according to the study as summarized in the article. The study also estimated that 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,” largely managing agents, rose from 10 percent in fall 2025 to 22 percent in spring 2026. It also reports that written standing instructions for agents grew from a few hundred to 8,500 over a year. Those figures describe Anthropic’s workforce and should not be treated as measurements of all workplaces.
Meyer’s proposed response is to count the work that follows an agent’s output, use agents to test ideas without turning every promising result into an ongoing project, set stopping conditions for recurring tasks and give a coordinating agent a limited attention budget. He suggests that such an agent remain silent by default and bring forward no more than three decisions a day, each with a recommendation, supporting evidence and the consequence of taking no action.
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
Human Review Sets the Real Cost
The article’s central point is that generating an answer quickly does not make the full task cheap. People still have to read, check and use what agents produce. A short request can create a large review burden if it produces a long report, or if several people must verify the result before acting.
That distinction matters for teams deciding whether to automate recurring work. Counting only agent runtime can make a workflow look efficient while leaving out the human time needed to manage it. Meyer’s approach would make that hidden work visible before a team adds more agents or alerts.
It also changes what a useful update looks like. A digest that reports activity may still demand attention without helping a person choose what to do. A recommendation paired with evidence and the cost of inaction gives the recipient a clearer basis for deciding whether a response is needed.
Products Worth Considering
Sony WH-1000XM4C Wireless Noise-Cancelling Headphones, Black
Bose QuietComfort Wireless Noise-Cancelling Over-Ear Headphones, Black
Sony WH-1000XM6 Wireless Noise-Canceling Over-Ear Headphones, Blue, Bundle
As an affiliate, we earn on qualifying purchases.
Anthropic’s Workforce as a Case Study
The article draws on a study Anthropic published on October 16, 2026 about its employees’ use of AI agents. Meyer notes that the data come from one company, where staff adopted the tools by choice, and says the study’s authors caution that Anthropic’s working practices may not transfer to other organizations. The statistics are therefore a case study, not a universal measure of agent use.
Meyer also cites examples from Anthropic customer stories to illustrate the difference between useful output and extra work. South Jersey chef Dan Ninerell attributed $20,000 in revenue over a month and a half to proposals he made with Claude; that is his own account, published by the vendor. Chicago-area painting contractor Tony Severino said he caught the model using a floor-area multiplier instead of actual wall measurements on an early bid and now asks it to show its work. These accounts are examples, not independent evaluations of the product.
“an army of Claudes, with every one asking for my attention.”
— An engineer quoted in Meyer’s article
How Far the Findings Apply
It is not clear from the article how closely Anthropic’s findings would match agent use in other companies, job types or levels of adoption. The study concerns one company’s workforce, and the source material says its authors warned that their employees chose to use the tools and that Anthropic’s practices may not generalize across the economy.
The article also presents Meyer’s four rules as practical recommendations, not as results from a controlled test showing they reduce workload. The supplied material does not report a measured impact from using a three-decision daily limit, default silence or end dates for recurring tasks. Nor does it specify how the study calculated all of its estimates, including the time required to read agent output.
Testing Attention Limits in Practice
Meyer’s recommendations amount to a set of workflow choices for people already using agents: ask who will review and act on each result, run small experiments without automatically keeping them, and give recurring jobs a clear end date and a condition for stopping when nothing changes. A team applying the advice would need to decide who owns review and what qualifies as a decision that warrants an alert.
The next useful evidence would be whether these limits actually reduce review time or improve decisions in workplaces beyond Anthropic. The source material does not identify a follow-up study or a planned company-wide test of Meyer’s approach, so the broader effect remains unmeasured.
Key Questions
What is the main point of Meyer’s article?
It argues that AI agents can generate more work for people when their output requires extensive reading, checking and follow-up. Meyer recommends designing agent workflows around the human attention they require.
What did Anthropic’s study report?
As described in Meyer’s article, employee-directed agents logged six times as many hours as their human directors by August 2026 and spent about 80 percent of session time waiting for human input. The findings concern Anthropic’s own employees, not the workforce as a whole.
Does the study prove that AI agents increase workload everywhere?
No. The article says the study examined one company and that its authors cautioned against assuming the results apply across the economy. It is evidence about Anthropic’s experience, not proof of a universal effect.
What does Meyer mean by asking agents for decisions, not updates?
He recommends that a coordinating agent stay silent by default and raise a small number of items that need a person’s judgment. Each should include a recommendation, evidence and the consequence of doing nothing, rather than a routine activity summary.
What remains untested?
The supplied article does not establish whether Meyer’s rules measurably reduce review time or improve decisions. It also does not show how well Anthropic’s findings apply to other workplaces.
Source: ThorstenMeyerAI.com
