Adding AI to Your Team Won't Fix It, Study Says

Companies everywhere are adding AI assistants (software that can carry out tasks on its own, from drafting emails to updating spreadsheets) to their teams. A new paper by researcher Zilong Wang argues that this alone won't make a team smarter. He built a computer simulation — think of it as a flight simulator for office routines — and ran it through six decision rules, three memory setups and three work environments, all under a fixed budget.

The main result: teams that kept a running record of what worked scored higher than teams that started fresh. On the study's own scoring scale, that gain was small but real (0.452 to 0.480). The bigger surprise involved teams that regularly stop and rethink their plans. When they kept the old evidence, they did far better; when they erased it each time, rethinking became nearly pointless. In plain terms, wiping your team's memory is expensive.

The catch is that memory can overstay its welcome. When the best way of working was reversed, hanging onto the old playbook forever delayed adaptation — a limited memory window eventually restored performance, though the switchover cost time. In further tests, the researcher equalized how much fresh information each approach gathered. The apparent advantage of constantly reorganizing shrank by two-thirds, suggesting much of the benefit came from gathering new information, not from replacing the process itself.

The practical takeaway: don't expect a shinier AI model or a new team lead to unstick a struggling team. Keep records, review them on a schedule, and when conditions genuinely change, be willing to clear out stale habits. Notably, replacing the rules themselves added no reliable benefit in any scenario tested. One honest caveat: this is a computer simulation, not a study of real employees, so treat it as a hypothesis about workplaces rather than proof.