AI can generate answers at extraordinary speed, but it’s quietly introducing a new constraint. Every prompt becomes a background process—and while the model works on a reply, we open another tab and start another prompt. Multiply this habit and our attention is scattered across many half-finished conversations.
As AI accelerates output, the real bottleneck is shifting from model capability to the human capacity to manage attention across many parallel AI threads.
Tools like Claude and ChatGPT promise to accelerate knowledge work. In many ways, they do. But after using them extensively, I’ve started to notice a subtle productivity tradeoff that isn’t talked about much.
The more detailed the instructions we give these systems, the longer they take to respond. Some prompts take a minute or more to process. That delay creates a strange gap in the workflow. While the model is generating a response, I start doing something else. I open another Claude window. I jump into ChatGPT. I check email or move across browser tabs.
AI is quietly becoming an attention fragmentation engine. I frequently find myself cognitively stalled and asking: Where was I? What was I doing? What threads are open and need to be checked?
The subtle shift
Instead of working through a single line of thought, we start running multiple parallel threads. Each AI response becomes a pending process in the background. Our attention splits across prompts, tools, and tabs.
Research on multitasking has warned about this dynamic for years. Studies consistently show that what feels like productivity is often just activity. Each context switch forces the brain to reload the original problem, creating a hidden cognitive cost. The result is busyness without real forward progress. Though, that research was done before AI.
AI can multitask much better than humans. That’s the whole point of agents. But, when every prompt becomes a waiting process, it becomes easy for humans to start five or six threads of work simultaneously. Parallelism increases throughput in machines, but in humans, it often increases friction.
Less friction increases thinking
In the pre-AI world, many high performers routinized the routine. Former President Barack Obama famously limited his wardrobe to reduce daily decisions. He had a closet full of the same suit. He never had to choose what to wear. The idea is not trivial. Every decision, no matter how small, draws from the same pool of mental capacity. Simplifying low-value choices frees cognitive capacity for higher-stakes thinking.
In knowledge work, boring systems are often the most productive systems. Teams that unify their tools, reduce platform switching, and standardize workflows spend less energy navigating work and more energy doing it.
The device we use also shapes how we think. People consistently choose larger screens like laptops or desktops for complex tasks because they provide more context visibility and reduce cognitive strain. Mobile devices, by contrast, constrain us to simpler, single-thread interactions. In some ways, that constraint forces focus.
AI tools sit somewhere in between. They offer enormous cognitive leverage, but their interface design still resembles chat windows scattered across tabs. The result is powerful capability layered on top of fragile attention.
Human judgment is still essential
AI does not remove the need for human judgment. In fact, it increases it. Someone still has to define the direction, structure the prompts, refine the output, validate the results, and ensure the work ultimately leads to the sought after outcomes.
Nothing the models produce is perfect. Every output still requires interpretation, adjustment, and orchestration inside a broader system of work.
This is why the smartest companies are not cutting people. They are investing in people who can own the end-to-end process. The emerging skill is not just using AI tools. It is directing them.
There is no AI that can orchestrate your business.
What is emerging instead is a new operating model. Humans increasingly act as conductors, coordinating multiple AI agents across different tasks. The organizations that win will not be the ones with the fewest employees. They will be the ones with the most capable people orchestrating the most agents.
Think of it less like replacing workers and more like scaling leverage.
A single person can now run dozens of parallel processes through AI. But those processes still need a human to guide them, align them with real objectives, and integrate them into a coherent outcome.
Which brings us back to the original challenge.
As AI increases the number of threads we can run at once, our ability to manage attention becomes the real bottleneck. The future of AI productivity will not just be about faster models or better prompts.
It will be about how well humans can orchestrate many intelligent systems without losing the thread of the work itself.
Louis Camassa is the director of product for Rithum.