AI (in its current form) is reshaping knowledge work by removing friction from production. Drafts arrive almost instantly, summaries appear before anyone has fully read the source material, and questions that once required an hour of back-and-forth can now produce polished answers in seconds.

On the surface, that looks like progress. In some ways, it is. But something important is disappearing: the resistance that used to sit between a question and a conclusion.

Someone once had to wrestle with the material, sort through conflicting inputs, and commit to a point of view before anything coherent could be shared. That process was inefficient, but it also created ownership over the thinking. Now the output can arrive fully formed, and teams often accept it as a reasonable starting point without much interrogation. Conversations move forward as if the groundwork has been done, even when no one in the room has actually done it themselves.

The result is a subtle but meaningful shift. Agreement may come faster, but it is often thinner. People align on language before they align on meaning, and speed can create the impression of clarity even when the substance underneath it is still unsettled.

For a long time, knowledge work was valued according to what it produced: a presentation, a report, a recommendation, a summary. If you could generate the artifact, you were doing the job. That assumption is weakening quickly, because the first version of almost anything can now be produced with very little effort. What remains difficult, and increasingly valuable, is deciding whether that version is any good, whether it is pointing in the right direction, and what should happen next.

Organizations have not fully caught up to that shift. Roles still emphasize deliverables, and performance is still measured through output, even as the more consequential work increasingly involves evaluating, questioning, and redirecting.

Many offices are designed for two modes of work: focused tasks and structured meetings. What they often fail to support are the less formal, more iterative interactions where judgment actually develops, where ideas are tested, revisited, and refined over time.

That gap shows up in our Workplace Experience Index. Employees report relatively strong support for individual work and formal collaboration, but meaningfully less support for the kinds of informal, evolving interactions where shared understanding tends to take shape.

AI helps organizations get to output faster. But faster is not the same thing as better.

Better still depends on the conditions that support discernment: hearing each other clearly, seeing the work, focusing without interruption, and moving fluidly between individual reflection and group sense-making.

If the environment only supports execution, teams will default to execution. If we want better decisions, better ideas, and deeper expertise, the workplace has to support the slower and less structured moments where those things are actually built.