Generative AI has done something strange in the knowledge work economy: it has dramatically reduced the cost of generating ideas.
Any reasonably capable professional with a chatbot can now produce a dozen plausible strategies, memos, product concepts or marketing plans before lunch. In some cases, AI also reduces the cost of execution, but not as much or as quickly. Implementing even one of those ideas still takes weeks, months or years.
The result is already being seen in every workplace: more initiatives than teams can carry out, more tools than anyone can learn, and more priorities than any reasonable person can keep in mind. Leaders continue to pile on new jobs because the cost of imagining new jobs has dropped to almost zero. But the cost of doing so has not been.
This creates a new management challenge: in an AI-saturated workplace, the bottleneck is no longer ideas. It is execution.
A cutting-edge genomics lab solved this problem about a decade ago, twice.
The Broad Institute lesson on doing less to do more
The Broad Institute, an MIT-Harvard biomedical research center, experienced one of the most rapid cost collapses in modern technological history. When the first human genome was sequenced in 2003, it took more than a decade and cost approximately $3 billion. Today, sequencing a human genome can take hours and cost less than $200.
That collapse created obvious opportunities, but also two separate crises at Broad.
The first was operational. As sequencing became faster, samples were moved through the process faster than downstream equipment could process them. Work piled up at bottlenecks. The lab became so overloaded that technicians began losing samples.
The solution was to move from a “push” system, where each stage sends work downward as quickly as possible, to a “pull” system, where each stage only accepts new work when it has capacity.
Then came a second crisis, which sounds a lot like the AI problem in the workplace.
Once sequencing itself became cheap and routine, Broad’s innovation team faced an explosion of ideas. New projects were constantly being started. Few were ever finished. As an MIT case study put it, the group was “losing the technological leadership position it had worked so hard to achieve.”
The solution was the same discipline applied to ideas.
The team created a visual map (literally sticky notes on the wall) of each active project and tracked where each one fell in the development funnel. The exercise revealed two things: some projects were redundant, and there were at least twice as many projects underway than the team could realistically handle.
They created a project funnel on the wall and added a “hopper” in front of it: a holding area where ideas waited until capacity opened up in the funnel.
In two years, the team reduced active projects by more than half and increased the number of projects that actually got done.
Why leaders keep adding work
Broad’s solution seems obvious in retrospect. It rarely happens in practice because humans are predisposed to addition.
A 2021 Nature A study led by researchers at the University of Virginia found that when people are asked to improve a design, document, or process, they automatically add instead of subtract.
In the workplace, that bias is compounded.
A new tool is implemented, but the old ones remain.
A new priority is announced, but the old priorities are not retired.
More meetings. More control panels. Longer strategy decks.
Most organizational complexity is the sediment of individually reasonable additions made without subtractions.
AI accelerates this dramatically.
It is now trivial to generate a 17th strategic priority, a fourth product line, or a third board. The obstacle is no longer imagination. They are the humans who are asked to execute.
What high-performing teams do differently
The companies that best adapt to this change are applying some version of Broad’s discipline.
Make active work visible
You can’t manage what you can’t see. Place each initiative in progress on a shared surface: a wall, a board, or a single document. Visibility forces classification.
Stop starting and start finishing.
In operations research, limiting work in progress is one of the simplest ways to improve performance. The new job waits until something else is finished.
Define “done” before you begin
Before you start a project, clearly define success.
Tony Fadell, who led the design of the iPod and co-founded Nest, told me that his most important advice for startup founders is to write the press release before starting the project. It forces teams to clarify priorities and define the finish line from the beginning.
None of this is about achieving less. What really matters is finishing the job.
In an AI-saturated economy, ideas are becoming a commodity. The advantage will be for organizations that can decide which ideas are worth implementing and which to ignore.
Adapted from INSIDE THE BOX: How limitations make us betterby David Epstein. Copyright © 2026 by David Epstein. Published by Riverhead Books, an imprint of Penguin Publishing Group, a division of Penguin Random House LLC.