What the research says in 2026
Five findings from peer-reviewed work this year, each with a decision it should change.
Psychological safety gained a clock
The best-backed theme on our radar produced its sharpest finding yet: in three-wave data, psychological safety lowers exhaustion over about two months, while exhaustion erodes safety within one. Burnout destroys the conditions for recovery faster than safety builds them, so waiting for a culture program to compound is a losing race. Decision: treat safety as infrastructure with a maintenance schedule, not an initiative.
AI productivity arrives with a coordination bill
The cleanest field result on AI at work shows real individual output gains paired with new coordination costs: the work of integrating, reviewing and reconciling machine-produced output lands on the team around the tool. Decision: budget the coordination work explicitly, or the gains evaporate into rework.
Meanwhile the youngest workers, across 1,485 participants and four generations, show the lowest trust in large language models. The assumption that Gen Z will drag adoption forward is backwards. Decision: retire it from every AI rollout plan.
Presence is not collaboration
Four leading researchers published the strongest version of both return-to-office cases side by side: co-presence genuinely enables tacit exchange and serendipity, and mandates genuinely cost autonomy and trigger reactance. The synthesis lands on design: decide which work needs co-presence and engineer for it, instead of buying badge swipes. And the employment deal keeps renegotiating in the details: 84 percent of hiring cost lands after the signature, in onboarding and ramp, which is where scarcity is actually won or lost.