Fragmented focus in theage of AI

March 25, 2026

This is post 4 in a series about being fast and good. Read part 1, part 2 and part 3.

The mathematical impact of distraction is sobering:

  • Individual distractions like emails or chat notifications (IM) create 8-10 minutes of diversion1.
  • The bigger cost is getting back on track. Resuming a task after diversion takes 10-15 minutes.
  • Resuming complex work is harder2, taking up to 2 hours in almost a third of cases.

Knowledge workers do adapt to distractions by working faster, but the cost is in compressed work product (shorter written output in the study), along with increased stress and frustration3.

If you want to be fast and good at your work, distractions are the enemy. But how do modern knowledge workers navigate distractions, especially in the age of AI?

A brief history of distraction and technology

This isn’t a new problem. In his excellent biography of Mark Twain4, Ron Chernow recounts the famous author’s struggle to overcome the distractions of social life in his own home. For a time he accomplished most of his annual writing work over the span of a few summer months, alone in a secluded hut outside of a family vacation home.

Technology isn’t the source of distraction; humans are good at finding diversion in almost any form. But technology is an amplifier of distraction’s potency and accessibility.

Nicholas Carr’s now classic 2008 article, Is Google Making us Stupid? 5, chronicled the distracting form factor of the modern web and its impact on attention and focus. Cal Newport famously eschews doom-scroll mediums like social platforms to focus on deep work6.

The modern, remote-heavy work environment itself is hostile to focus in the default state. Chat platforms like Slack and Teams are theoretically asynchronous, but in practice reward synchronous responses and are designed for notifications by default.

Fragmented focus in the age of AI

AI is fascinating because of its ability to both eliminate distractions and create new contexts for them.

Slack’s Recap feature uses AI to summarize conversations across channels, saving users hours of manually combing through messages. This type of proactive consolidation of disparate information dramatically reduces the need for notifications, bookmarks and unread conversation hoarding. Similar tools exist for email, issue tracking and version-controlled repos.

When AI removes distraction, it happens in the context of existing workflows and tools, but LLMs have also created entirely new ways of working, which creates new forms of distraction.

Context switching: more work streams on faster cycles

Earlier in this series I argued that AI can be dramatic accelerator for people who are fast and good at their work.

Practically, this means that:

  1. You can do more work in the form of additional projects
  2. You can accomplish many of those projects on a compressed timeline (relative to the non-AI baseline)

In a vacuum, an individual worker can execute those projects in prioritized sequence, faster, resulting in significantly higher output.

On the ground in real companies, projects are coordinated across business functions, which shape them through competing priorities, and are executed in parallel.

For teams extending existing output with AI (as opposed to just adding headcount), the result is far more context switching.

Context switching doesn’t have to be the same as distraction, but the nature of the experience is often indistinguishable as project timelines and their deadlines compress into smaller windows.

If you normally work on 3 projects in a week, AI augmentation may increase your capacity to 5. That’s a 60% increase in project load, but it’s not just work. Projects require stakeholder management, meetings, due-date coordination and the inevitable interruptions that come with them. Resumption costs can pile up quickly.

Unpredictable down time waiting on model responses

As AI models increasingly take on complex work, round-trip time on a request can take several minutes, or even hours, depending on the task.

Because response time is variable, this creates an interesting challenge for workers using AI: how do you use the downtime waiting on the model to respond?

If you have notifications turned on, switching to another task guarantees interruption when the model notifies you that the task is complete.

This is a strange new cocktail of efficiency, but on an unpredictable timeline, in a form factor that makes interruption difficult to avoid.

Coordinating multiple agents

The challenge of downtime use is exacerbated by the ability to run multiple agents at the same time, all working on different projects, all with unpredictable response timelines.

Being fast and good requires focus

Even with the efficiency of AI, interruptions put a quantifiable ceiling on how quickly can you complete work.

The more pernicious problem is that unmitigated distractions limit you from building skill over time (becoming really good).

We all know that our best work happens when we gan give it our full attention for extended periods. That same process of deep focus is also what sharpens our mental tools and builds the additional skills required to become better at a craft.

So, what to do?

AI distractions aren't new, they are modern manifestations of the age-old problem, and the answers are the same today as they were for Mark Twain.

Time-box work and turn off notifications

People who produce the best work are often most protective of their calendars, whether or not that’s apparent to their coworkers. They block time for focused work and eliminate interruptions by turning notifications off (including pings from agents that completed a job).

This can be harder than it seems because distractions are dopamine. Focus time is a muscle, not a switch, and the more you use it, the more valuable it will become.

Single-thread intentionally

There are contexts for running multiple AI chats or agents simultaneously, but just because you can doesn’t mean you should.

Finishing a single task, especially a mentally taxing one, creates momentum through accomplishment and decreases the perceived effort of future tasks7. When using AI, this means choosing to stay in the same mental frame on the same project, even when you’re waiting on a model to respond.

Produce more than you consume

If you want to be fast and good, one of the best litmus tests for problematic distraction is assessing whether you produce more than you consume, without extra hours as a crutch.

If you’re producing at least as much work product as you consume, you’ll get faster and better over time.

Join almost 3,000 subscribers via Substack to get notified about new posts.

Footnotes

  1. The field study Disruption and Recovery of Computing Tasks quantifies the time cost of distractions. ↩

  2. A Diary Study of Task Switching and Interruptions tracks and analyzes the challenge of resuming tasks after distractions occur. ↩

  3. The Cost of Interrupted Work: More Speed and Stress uncovers the intangible negative impact of distraction. ↩

  4. Ron Chernow’s biography of Mark Twain is a fascinating read and details multiple aspects of Twains writing process. ↩

  5. I remember reading Is Google Making us Stupid as an intern working at what was then considered a high-tech marketing agency. Carr went on to write The Shallows, an expansion of the work in the original article. ↩

  6. Deep Work came out in 2015. The principles are timeless, but feel more relevant today than they did a decade ago. ↩

  7. James Somers wrote a helpful article on the benefits of being able to work quickly (which helped plant the seed for this series). He argues that “the obvious benefit to working quickly is that you'll finish more stuff per unit time. But there's more to it than that. If you work quickly, the cost of doing something new will seem lower in your mind. So you'll be inclined to do more.” ↩