The Most Dangerous Thing AI Can Remove Is Friction
By Amanda M. Main, Ph.D.
The most interesting question in AI right now is not what it can do; it is what we should have the sense to leave alone.
Organizations are currently behaving as if every slow, annoying, inefficient part of work is a problem waiting for a bot. Draft the email. Summarize the meeting. Rank the applicants. Write the performance review. Answer the customer. Generate the proposal. Remove the pause. Remove the handoff. Remove the waiting. Remove the part where a human being has to sit with something for more than twelve seconds.
And some of that is wonderful, truly. No one should need to painstakingly hand-format a spreadsheet to prove their professional worth. But somewhere along the way, we started treating all friction as failure and that is a mistake.
Some friction is just nonsense like the form that asks for the same information three times, the approval chain that exists because no one has had the courage to delete Gary from the workflow or the meeting that could have been an email.
But some friction is doing real work. The pause before replying to an angry client may be the reason the relationship survives. The second read of a policy decision may be where someone notices the quiet legal or ethical problem. The awkward silence in a meeting may be the only opening a junior employee gets to say, “I don’t think this is right.” The boring task between two demanding conversations may be the tiny mental reset that keeps a person going into social burnout.
That is the part many leaders are missing. They see inefficiency and assume it has no value, but work is not just a series of outputs. It is a human system full of judgment, trust, fatigue, power, timing, and risk. Make everything faster and you may not get a smarter organization, you may just get a more fragile one.
The strategic move is not using AI everywhere you can; it’s admitting that some parts of work get worse when they get faster. This is especially true in moments with consequences: hiring, promotion, conflict, customer recovery, ethical calls, strategic decisions, or anything involving people who have less power than the person making the decision. AI can absolutely help in those spaces by summarizing, organizing, drafting, comparing, and surfacing patterns and in those instances it should be deployed.
But it should not become a trapdoor where responsibility disappears. A polished recommendation is not the same thing as a good decision; a confident summary is not the same thing as understanding; and a faster process is not automatically a fairer one. In fact, AI makes some mistakes more dangerous because it makes them look better. Bad thinking used to show up with typos and weird formatting but now it comes in a clean paragraph with a subject line and three bullet points.
This all means that discernment is about to become one of the most valuable leadership skills in business. Not technical fluency alone, nor blind enthusiasm. Discernment.
Knowing where speed helps and where it harms; which tasks drain people and which ones develop them; when a pause is waste and when it is wisdom; and when AI should drive, when it should assist, and when it should sit quietly in the back seat.
The next competitive advantage is not going to belong to the companies that automate the most. It is going to belong to the companies that can tell the difference between useless friction and necessary friction. Because AI makes speed cheap. Judgment is still expensive. And it should be.
Connect With Amanda
Comments