🔍 Read the full analysis: Stop Asking AI Agents For Status; Ask For A Decision on ThorstenMeyerAI.com
Get business pricing on tech for your team
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
A report published by ThorstenMeyerAI.com on October 27 argues that AI agents should be judged by the human work their output creates, not just by how quickly they complete tasks. It points to Anthropic’s October 16 study, which found long agent wait times and heavy output-review demands among its own employees, and recommends limiting updates to a few actionable decisions.
A report published October 27 by ThorstenMeyerAI.com argues that people coordinating AI agents should ask for decisions rather than routine status updates, citing an Anthropic study that found agents in employee-directed work spent about 80 percent of session time waiting for human input. The recommendation addresses a practical constraint as organizations deploy more agents: the time people need to review their output and decide what to do with it.
The report draws on Anthropic’s study of its own workforce, published October 16, 2026. It says that by August 2026, agents directed by Anthropic employees logged six times as many hours as the employees directing them. The report also cites the study’s estimate that reading all agent output would take non-technical employees about half their working time. These figures describe Anthropic’s staff and the study’s estimates; they are not measurements of every workplace.
Thorsten Meyer proposes four operating rules: count the human work that follows an agent’s output; use agents for small experiments but keep only those that prove useful; set an end date and a stopping condition for recurring tasks; and give a coordinating agent an explicit attention budget. Under the last rule, the coordinator defaults to silence and raises no more than three decisions a day, each with a recommendation, supporting evidence and an explanation of what happens if no one acts.
The report’s central distinction is between the low cost of generating material and the continuing cost of reading, checking and acting on it. It warns that a quick automated draft can still create substantial work if several people must review it, or if its conclusions need careful verification. The proposed approach is to make agent output easier to evaluate and connect it to a decision someone is prepared to make.
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
Managing the Human Review Load
The advice matters because an agent can increase the amount of work available without reducing the time needed to judge that work. If a system sends frequent updates, drafts and suggestions, a person still has to decide which are relevant, check important claims and follow through. At a large scale, the review burden can weaken the time-saving case for automation.
A decision-focused update could reduce interruption and make responsibility clearer: the agent states what it recommends, why, and what the consequence of inaction would be. That does not remove the need for human oversight. It aims to direct that oversight toward choices with consequences instead of asking people to monitor every step. The report presents this as a practical operating proposal, not as a result proven across organizations.
Products Worth Considering
LG 34GX90SA-W 34-inch Ultragear WQHD OLED Curved Gaming Monitor
Samsung 32-Inch Odyssey OLED G8 (G80SH) 4K Gaming Monitor
Samsung 27-Inch Odyssey QD-OLED G8 4K 240Hz Gaming Monitor
As an affiliate, we earn on qualifying purchases.
The Anthropic Study Behind the Advice
The source report describes Anthropic’s research as an examination of how its own employees used agents, and says the study cautioned against treating its findings as universal. Anthropic’s staff adopted the tools by choice, and the company’s working practices may not match those of other industries. The reported figures should therefore be read as company-specific evidence, rather than a workforce-wide benchmark.
The report also points to a change in agent management at Anthropic: the share of sessions devoted to “meta-work,” described as mostly managing agents, rose from 10 percent in fall 2025 to 22 percent in spring 2026. It says written standing instructions for agents grew from a few hundred to 8,500 over a year. These measures suggest that coordinating agents can itself become a significant part of the work, though they do not establish how much time or effort other organizations will spend.
“The bottleneck is your attention, not the agent.”
— Thorsten Meyer, author of the October 27 report
How Far the Findings Apply
The available source material does not establish whether other companies see the same agent wait times, review demands or growth in coordination work. Anthropic’s study concerns its own employees, while the report itself says the company’s voluntary adoption and working practices limit how broadly the findings can be applied. The reported estimates also do not show whether agent use improved productivity overall after accounting for review and management time.
It is also unclear how widely the proposed limits—such as three decisions a day or a default of silence—have been tested. The report offers them as operating rules based on Meyer’s experience, not as independently evaluated standards. The source material provides no comparative trial showing that decision-oriented updates outperform other approaches.
Testing a Lower-Noise Workflow
The immediate next step for readers adopting the proposal is to apply its checks to individual agent tasks: identify who will read the result, who will act on it, and whether the resulting work is worth taking on. Recurring tasks can also be given a defined end date, a “nothing new” stop condition and a pause if their output is no longer being read.
Whether these practices save time at scale remains an open question. A broader comparison across organizations would be needed to establish how common the attention burden is and whether decision-limited reporting measurably reduces it. Until then, the report’s recommendations are best understood as practical guidance informed by a company-specific study, not a settled formula for managing every AI agent.
Key Questions
What is the report’s main recommendation?
It recommends asking agents to surface decisions that need attention rather than sending frequent routine status updates. A proposed update includes a recommendation, evidence and the consequence of doing nothing.
What did Anthropic’s study report about agent waiting?
The report says Anthropic’s study found that agents in employee-directed sessions spent about 80 percent of session time waiting for human input. The figure describes the company’s study, not all AI agent use.
Does the study prove AI agents reduce productivity?
No. The cited figures describe agent activity, waiting and estimated reading demands at Anthropic. They do not, by themselves, establish whether agent use raised or lowered overall productivity.
What does an attention budget mean in this proposal?
It means setting a limit on how often a coordinating agent interrupts a person. Meyer suggests a default of silence and a maximum of three decision requests a day, but the source does not report that this limit has been validated across workplaces.
Source: ThorstenMeyerAI.com
