I not too long ago seen a paradox amongst a crew of builders. With AI, engineers began writing code sooner and getting solutions in seconds, but in addition they reported feeling extra exhausted than earlier than.
AI hasn’t truly lowered the quantity of labor that must be achieved. As an alternative, it has basically modified its nature. We are able to now run a number of duties in parallel and understand this as productivity. Up to some extent, it’s. However finally, managing instruments and consistently switching between them turns into extra draining than performing the unique duties themselves. In some circumstances, it even slows down the method of discovering an answer.
I’ve been managing developer groups for over 15 years, and I’ve spent the previous yr making an attempt to know why AI instruments—designed to make work simpler—generally have the other impact. Listed below are the causes behind this phenomenon and what we are able to do about it.
WHERE THE FATIGUE COMES FROM WHEN AI IS DOING PART OF THE WORK
Take a developer’s workflow for example. Prior to now, when confronted with a fancy downside, builders would search Google, use Stack Overflow (earlier than ChatGPT arrived), and ask colleagues for assist. Every step and choice was separated by reflection time.
Now, they begin utilizing Cursor or GitHub Copilot—AI instruments that counsel code in actual time. The trail to a solution will get shorter. However as an alternative of looking out, they’re now engaged in steady analysis of AI strategies: Settle for the autocomplete or reject it, rewrite the immediate or regenerate the output. Dozens of micro-decisions with no pauses between them.
Every of those carries a cognitive “value.” Even the smallest alternative calls for consideration and psychological effort. The extra choices an individual makes each day, the more serious the standard of every subsequent one. This occurs due to what psychologists name decision fatigue. AI has amplified this downside by dramatically growing the variety of choices an individual should make whereas finishing a single job.
Researchers at Boston Consulting Group (BCG) surveyed almost 1,500 U.S.-based staff. They discovered that 14% of people that use AI at work, needing a excessive quantity oversight, expertise “AI mind fry”—a sense of psychological fog and an incapability to focus. And this has penalties: Employees experiencing it usually tend to contemplate altering jobs and make extra errors.
MORE TOOLS DON’T IMPROVE PRODUCTIVITY
I’ve seen it repeatedly: Managers start implementing AI with the identical query: How can we use these instruments to assist the crew get extra achieved? Then they begin including AI providers.
One or two AI instruments genuinely do enhance productiveness, per BCG, however at three instruments, productiveness peaks. With the fourth, it drops. Every new instrument means new settings, prompts, and workflows. In some unspecified time in the future, the crew spends extra effort managing instruments than doing the precise work.
The employee stops being the doer and turns into the one checking, evaluating, and selecting. In the meantime, individuals stay satisfied that AI makes them sooner. However in line with METR data, the other was true: Skilled builders utilizing AI instruments truly labored extra slowly—even whereas believing their job completion time decreased by almost 1 / 4.
There’s one other paradox right here. Even when AI genuinely hurries up work, individuals don’t use that point to relaxation. They tackle extra duties. This was found by researchers at UC Berkeley’s enterprise faculty, who spent eight months observing staff at an American tech firm to know how AI utilization affected their work habits.
At first, staff felt energized; their productiveness soared. However over time, the workday quietly stretched longer—a immediate throughout lunch, one final question earlier than leaving the workplace—whereas the variety of breaks decreased. Nobody demanded they work extra, however that’s precisely what occurred. Later, those self same staff admitted they had been exhausted. The researchers known as this “workload creep”—a gradual enhance in workload that accumulates unnoticed till fatigue begins affecting choice high quality.
SHOULD WE ABANDON AI TOOLS?
I don’t assume abandoning AI is the reply. I’m satisfied the issue isn’t the know-how however how we use it and our targets. Listed below are 5 issues that, in my expertise, will help implement AI with out burning out your crew:
1. Begin by rethinking your workflows. Earlier than introducing any AI instrument right into a course of, start with a query: Which duties require human consideration, and which may be automated with out sacrificing high quality? The method of “implementing AI in each course of” isn’t a technique—it’s a quick observe to breaking what already works.
2. Give managers a number one position in AI adoption. In groups the place the supervisor personally helps individuals be taught AI instruments, cognitive fatigue amongst staff is decrease, in line with the BCG research. A supervisor who really understands how AI works can set a wholesome tempo for utilizing these instruments and stop the crew from drowning in experiments.
3. Set up guidelines for working with AI. The Berkeley researchers known as this “AI apply”—agreements about how the crew engages with AI. These would possibly embody a brief pause earlier than necessary choices, sequential execution as an alternative of multitasking, and time for dialogue and collective reflection. One in every of our crew leads, for instance, encourages juniors to argue with AI extra usually.
4. Monitor cognitive load. We often conduct crew well being checks—monitoring productiveness, engagement, and stress ranges. However I’ve realized that’s now not sufficient. In our new actuality, cognitive load must turn into a separate metric. You can begin with a couple of questions: What number of AI instruments is somebody utilizing, has AI simplified their work or elevated its quantity, and the way does the worker really feel on the finish of the day?
5. Clarify the explanations behind adjustments to your crew. Folks may be skeptical of AI due to uncertainty. If an organization doesn’t clarify why it’s introducing new instruments, staff begin decoding it themselves. Against this, gaining an understanding that the corporate values stability—relatively than merely wanting extra output for a similar value—reduces psychological fatigue by 28%, per the BCG analysis. That is precisely the method I observe with my 100-person software program crew: transparency.
The important thing query isn’t “how will we use AI” however “why?” Begin with the aim of liberating individuals from routine duties. Enhancing choice high quality will yield totally different outcomes than measuring implementation success in tokens or strains of code.
Illia Smoliienko is chief software program officer of Waites.
