Anthropic's Employees Spend 20% More Time in Meetings After AI Adoption

The AI company where AI increased meetings

Anthropic, one of the most prominent enterprise AI companies, analyzed how AI was changing its own workplace. The result was surprising: employees were spending 20% more time in meetings.

That finding cuts directly against the central promise of workplace AI. The pitch is that intelligent assistants will eliminate coordination overhead, summarize discussions, automate routine tasks, and give employees more time for focused work. At Anthropic, however, some work still required direct human interaction — and the company's rapid AI-driven operating model appears to have generated additional coordination demands rather than reducing them.

The stranger detail is what employees reported about their AI tools. They described "nagging chatbots" that created more work instead of simply removing it. Bots could remind workers about tasks, chase updates, or prompt follow-ups. But every prompt became another item to process, answer, verify, or delegate. The automation made it easier to initiate work while increasing the number of things people were asked to respond to.

Why this matters

This is not a typical enterprise customer experimenting cautiously with new software. Anthropic is an AI company whose business depends on understanding how advanced models affect knowledge work. Its internal experience offers an early glimpse of what happens when the people building the technology are also among its earliest and most intensive users.

The broader B2B lesson is that productivity is not the same as activity reduction. If AI makes it easier for every employee, team, and software agent to request information, schedule a discussion, generate a review, or assign a follow-up, the organization may end up with more coordination despite better tools.

That creates a workplace paradox. AI can reduce the effort required to do individual tasks while expanding the volume of collaboration around those tasks. An employee might complete a report faster with AI assistance — then spend the saved time in three meetings discussing that report, responding to automated check-ins about it, and updating systems that track its progress.

The attention economy inside AI companies

Anthropic's experience suggests that employees may not experience AI mainly as an autonomous assistant. They may experience it as a new class of digital manager — always available, difficult to ignore, and constantly prompting them to update systems or complete the next action. The technology does not need formal authority to influence behavior. Repeated reminders create a culture of continuous responsiveness.

This is different from the familiar enterprise software complaint about too many dashboards or complicated interfaces. The issue is not that the tools are hard to use. The issue is that they work exactly as designed — by making coordination frictionless, they make it constant.

Every meeting becomes easier to schedule. Every request becomes easier to send. Every follow-up becomes easier to automate. The cumulative effect is an organization where the stream of coordination requests grows faster than the time saved by individual efficiency gains.

What comes next

The conventional enterprise software narrative assumes that every new layer of automation naturally produces fewer meetings and leaner workflows. Anthropic's internal data suggests a more complicated possibility: AI may be optimizing individual productivity while worsening organizational congestion.

That raises questions about the next phase of workplace AI adoption. If the company building frontier AI is seeing 20% more meeting time and complaints about nagging bots, what happens when these tools reach organizations with less technical sophistication and less tolerance for experimental workflows?

The next enterprise software backlash may not target automation itself. It may target the relentless prompts, alerts, and requests that automation enables. The issue will not be whether AI can complete tasks faster. The issue will be whether anyone can hear themselves think amid the notifications.

Anthropic's experience is a useful early warning. The risk is not that AI will fail to make work more efficient. The risk is that AI will make work more coordinated, more monitored, and more interrupt-driven — even at the companies that understand the technology best.

The 20% increase in meeting time is not a bug in Anthropic's implementation. It may be a feature of AI-native work itself.