How AI breaks flow, fragments attention, and quietly changes how product teams think
Monday morning I sat down to write about AI-induced distraction. Then, while waiting for ChatGPT to finish a response to a proposed outline, I checked Slack. Next response? My phone. By the time the output appeared, I’d forgotten the critique I wanted to make.
It can happen to me two to three times an hour. Some detours last five minutes. Across a workday, I estimate I lose an hour or more to context-switching—not because I lack discipline, but because AI’s variable latency creates 10-second gaps my brain can’t ignore.
I’m not alone. When AI takes 10 seconds, 30 seconds, or two minutes to respond, the herky-jerky rhythm invites distraction. The problem isn’t individual willpower—it’s a mismatch between how the tools behave and what humans can actually sustain.
And if I’m not the only one having this problem, then it isn’t a matter of me becoming more focused or mindful or present, but actually that there’s a structural issue here around how we work with AI. The way AI is currently designed encourages us to lose focus and flow. This has influence not only on our own individual productivity but on how we function on product teams more broadly.
Why “just focus harder” doesn’t work
The common prescription is mindfulness: stay present, resist distraction, build better habits. But that misses the structural problem.
At this point, we all know the struggles of waiting for AI to respond. It’s variable. Sometimes we get an immediate response, and sometimes we’re waiting for a minute or two for it to run. And we never know how long it’s going to be.
There’s a long-standing body of UX research on how long humans are willing to wait for the computer to get back to them. “It’s incredibly taxing to stay focused and alert in the absence of activity, but users can keep their attention on the goal for about 10 seconds.”
As one commenter shared on Reddit:
Whatever you choose to do in that couple of minutes of waiting, is something that breaks the flow. It takes time to go back to what you were dealing with and it’s so frustrating. Can’t have a single hour of focus since vibecoding started. –aleciak9669
After 10 seconds, attention drifts. It’s not a character flaw, it’s just how we’re wired. AI’s variable latency creates dozens of these micro-gaps daily. Willpower isn’t the solution; better structure is.
So we’ve got a conflict between how our tools are behaving with us and what we can actually do with them with regards to mental stamina.
How people cope while waiting (the modalities)
I asked around to see how people handle AI waits. The strategies fell into clear patterns—some effective, most not.
Broadly, everything falls into two categories:
- More mindfulness
- Multitasking, multi-threading
On a typical day, I get distracted from my AI two to three times an hour; some of those detours stretch five minutes or more. Across a full workday, I worry that adds up to roughly an hour lost to context switching instead of staying with the thread while I wait.
Here’s what people reported trying on Reddit, Hacker News, and Lenny’s Community:
Parallelization: “I vibe code multiple projects at once and rotate between them.” — thehen. Others on Slack mentioned starting another prompt or opening a different AI tool while waiting.
Micro-task switching: “I often end up spending 5–10 minutes on YouTube while waiting.” Others mentioned catching themselves checking Slack or email during short pauses. This is the bucket I fall into most often.
Scenario priming: “I use AI to plan my next prompt as it’s executing the last prompt.” — Happy_Acanthisitta92.
Embodied reset: Lenny’s community members mentioned standing up, stretching, or grabbing water to avoid sitting there getting irritated.
Mindful waiting: TackyFish shared “a slightly philosophical take on this – what’s actually wrong with sitting idle for those moments? This mindset may save time in real. [sic] Since you don’t fragment attention, you don’t lose the thread, and you engage with the response immediately and more thoughtfully.”
Tool orchestration: “Email is a better interface for long-running AI tasks than chat.” — Rahul Gupta-Waksal. Sometimes it’s better to not be explicitly waiting for the AI to finish typing, but instead change the interaction overall.
Each of us needs to figure out which modalities of waiting work better for our own brains. There is nothing wrong with any of these options. Underlying all of this is a broader conversation about how we view the necessity to be productive with every minute of every day and whether that’s actually healthy or productive.
But until the tools provide us with a more comfortable interaction that allows us to feel like we really are in an active conversation with another thinking partner, we are all going to have to figure out how we deal with these moments of waiting.
I now decide upfront: will I plan the next prompt, set a notification for longer tasks, or take a physical break? Having a default prevents drifting to Slack on autopilot. It’s not about perfect focus—it’s about giving your attention somewhere productive to sit while AI thinks.
Why this is a leadership problem
Individual distraction compounds into team-level costs. When everyone’s reasoning happens in private AI chats, then gets disrupted by micro-switches, product teams end up with:
- Thinner meetings: People half-listen while prompting in parallel. Decisions get vague agreement, not real alignment.
- Degraded reasoning: When I’m coming back to a topic after context switching, I find it harder to really read the AI output in enough detail to thoroughly think it through.
- Invisible work: Leaders can’t tell if silence means thinking or distraction. Engagement becomes hard to read.
The leader’s job is not to police focus. Our job is to coach people through a new form of cognitive load that fragments both individual thought and shared understanding.
How leaders can manage the distraction
As an operations person, my instinct is to build systems and tooling to handle fragmented attention. But the tools aren’t there yet, and as long as AI usage remains largely individual, this isn’t really an ops problem to solve.
Here’s what I’ve come to realize: for now, this is a coaching challenge, not a tooling one. Especially when working remotely, we need to shift how we interpret team behavior and watch for specific red flags.
Silence doesn’t mean thinking anymore
Silence used to signal reflection, processing, or careful consideration. When we’re working with AI, silence increasingly means attention has gone elsewhere.
In meetings, silence might mean someone is prompting AI in parallel, waiting on output, or already context-switched away. The danger isn’t the multitasking itself—it’s misreading engagement. You think everyone’s aligned when they’re only half-listening. Decisions get made without everyone actually processing them together.
Meetings start to feel thinner: fewer strong reactions, more vague agreement, less productive disagreement. People are physically present but cognitively split. The meeting still happens, but the shared cognitive moment is weaker.
Speed isn’t the same as good reasoning
When AI outputs arrive quickly and continuously, it’s easy to mistake speed for efficiency. But when attention is fragmented—jumping between AI and other work—the quality of thought quietly drops.
AI outputs get skimmed rather than interrogated. Subtle caveats get missed. Earlier assumptions fade as conversations stretch across hours. You end up responding to whatever AI just said rather than thinking through a coherent strategy. You’re moving fast, but you’ve stopped thinking deeply.
This is particularly risky because AI responses sound confident and complete. Errors are plausible rather than obvious. The burden of critical thinking shifts more heavily onto humans.
The cost shows up later: decisions that feel “off” but are hard to diagnose, missed implications no one can trace back, a nagging sense that something important slipped through. Without real attention, speed can make shallow thinking worse just as easily as it boosts productivity.
Name it, don’t normalize it silently
As team leads, we need to make sure people are thinking through this problem and handling it well. We can support them—and get better at it ourselves.
In your next 1:1, ask whether waiting for AI breaks their focus. Acknowledge that this is genuinely hard. Share your own struggle briefly and compare strategies without judgment.
This is time management coaching for the AI age. It can’t be a “you should stay focused” lecture. Approach it with empathy. This is hard for all of us—not something to feel shame about, but something worth working on together.
Start with intention, not willpower
Writing this article took twenty minutes longer than it should have. I got distracted while waiting for AI responses, but also just because my brain does that. When Slack messages come in from client work, I chase the shiny distraction rather than stay focused.
But thinking deliberately about this challenge has helped. I realized I’d been defaulting to distraction—messages, tasks, my phone, whatever was closest.
Now I set an intention before I start any AI conversation. Which modality do I want to spend the next hour in? Will I plan the next prompt while this one runs? Set a notification for longer tasks? Take a physical break?
Having a default prevents drifting to Slack on autopilot. It’s not about perfect focus—it’s about giving your attention somewhere productive to sit while AI thinks.
This is also where to start when coaching your team. Don’t expect willpower to overcome structural problems. Help people recognize the pattern, name their default distraction, and choose a better one.