There’s an old saying that necessity is the mother of invention. When there’s no easy way out, we’re forced to think differently and come up with something new. AI offers a shortcut around that. It gives instant answers with almost no effort needed.
That’s great for productivity in the short term, but it comes with a hidden cost. It quietly takes away the pressure that pushes people, and whole organizations, to get better at solving problems.
Is that a problem? Not straight away. Spending less time on a task means lower costs and more capacity for other things. But over time, it risks wearing down the skills a business actually needs to keep coming up with new ideas. We call this the “productivity trap”: you get faster in the short run, but slowly lose the ability to think for yourself.
Latest Videos FromTechRadar
Associate Professor of Strategy and Behavioural Science at Imperial College London and Department Editor.
We’ve spent time studying how organizations learn. We built a model looking at how people decide whether to work through a problem on their own or just reuse a solution someone else already found.
The pattern is clear. When “good enough” answers are free and instantly available, people reuse them more and explore less. Teams start settling on the same handful of approaches. Output goes up, but fresh thinking quietly dries up.
This isn’t just a theory. A well-known 2015 study looked at computational biologists and found that when they could easily see each other’s half-finished work, they spent more time polishing what already existed and less time trying new approaches. The range of ideas being explored shrank.
We’ve seen something similar in our own workplaces – with weekly sessions set up to share new AI tools were well attended, but over time they produced fewer genuinely new ideas. A small group of people kept doing the exploring. Everyone else just waited to pick up whatever they found. For business leaders, the lesson is fairly simple. AI changes how people learn at work.
If a decent strategy memo or risk assessment can be produced in minutes, fewer people will bother doing the slower work that actually builds understanding, like talking to customers, checking the numbers themselves, or testing their own assumptions.
Left unchecked, this leaves you with a team that’s good at getting answers out of AI but not very good at judging whether those answers are actually right.
Adding some friction back in
One option is to slow down how much AI your team uses. But that gives up real efficiency for a benefit that’s hard to measure, which is a tough sell to any leadership team focused on this quarter’s numbers. A better approach is what we call “strategic friction”: a small, deliberate step you add between an employee and an AI generated answer, so they have to think for themselves before they get it.
This isn’t about adding pointless red tape. It’s about making sure that before someone benefits from a ready-made AI answer, they’ve spent enough time on the problem to actually understand it.
Slightly counterintuitively, our research suggests this doesn’t slow people down overall. It speeds them up, because someone who has already wrestled with a problem is much better placed to spot when an AI answer is wrong, and to make it better. There are a few practical ways to do this, from simple rule changes to more involved system design.
The easiest option is asking for proof of independent effort first. Before someone is allowed to use AI to draft a market assessment, ask them to write up their own rough version first, even if it’s incomplete, showing what they already know and where they got stuck. This needs no new tools or software, just a change in how you work.
A step further is designing your AI tools to ask questions before giving answers. If someone asks for a competitor analysis, the tool could first ask them for a few things they already know or have noticed. Then it builds its answer around that. This gets people thinking before they receive an answer, and it usually makes the AI’s output better too, since it’s working from fresh information instead of generic guesses.
The most involved option is building tools that only unlock once someone has put in their own input. For example, someone might need to upload their own starting assumptions before they’re allowed to generate a risk assessment. It takes more work to set up, but it makes sure every AI output is a genuine team effort between the person and the tool, not just something handed over wholesale.
Not every task needs this. For simple, low stakes jobs, like formatting a slide or summarizing some notes, speed is all that matters, so let AI do its thing. Save the friction for the work where you actually need people to think, not just churn things out.
Spotting the builders
There’s a hiring angle here too. In a workplace where everyone has access to AI, the person who gives the “right answer” is no longer that impressive. What matters more is how they got there.
Look for people who use AI selectively and question what it gives them, rather than people who just accept whatever it produces and pass it on. Simply watching how someone uses AI, whether they push back on it or just go along with it, tells you a lot in an interview or a performance review.
AI is going to keep making businesses faster. The harder job for leaders is making sure that speed doesn’t quietly wear away the judgement and creativity the business relies on. A bit of well-placed friction, built into your tools, your workflows and how you hire, is what keeps a business coming up with new ideas instead of just churning out answers.
We’ve featured the best AI website builder.
This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.
The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit
