The variance is collapsing
The first thing to acknowledge is that the manager role carries an enormous amount of weight. Gallup's 2015 research found that managers account for 70% of the variance in team engagement scores. That is a between-team comparative number, not a claim that managers cause 70% of engagement. The remaining variance lives in pay, fit, policy, and macro conditions. But within an organization, the team-to-team gap is mostly the manager.
The second thing to acknowledge is that the variance is now collapsing in the wrong direction. Gallup's 2025 State of the Global Workplace reports global engagement dropping from 23% to 21% in 2024, the first drop since 2020. The drop is not driven by individual contributors, who held flat at 18%. It is driven by managers, whose engagement fell from 30% to 27%. The under-35 manager cohort dropped five points. Women managers dropped seven. Gallup put the productivity cost at $438B.
Read the two numbers together. Managers carry the variance. The carriers are buckling. The cohort the rest of the team depends on for trust, decisions, and tempo is the cohort whose own engagement is falling fastest. The role itself is breaking, and the conditions are unlikely to improve through better intentions.
The proximity tax
The most useful frame for what changed is the proximity tax. In a co-located org, a great deal of chemistry work happened without anyone scheduling it. Hallway exposure produced ambient trust. Shared lunch produced cross-group ties. Whiteboard sessions produced shared mental models. Standup was an excuse to read the room. None of that was a deliberate ritual. It was the default behavior of bodies in space.
Hybrid taxes that default. Microsoft and LinkedIn's 2024 Work Trend Index found 49% of hybrid managers struggle to trust their teams compared with 36% of in-person managers, with 54% reporting a visibility gap against 38% in-person. Microsoft's September 2022 Pulse captured what happens next: 85% of leaders said the shift to hybrid made it harder to be confident their employees were being productive, only 12% had full confidence in their team's productivity, while 87% of employees self-reported as productive. The trust collapse is asymmetric. Managers do not believe what their people report.
The networks themselves are also fraying. Longshi Yang and colleagues, in a 2022 Nature Human Behaviour paper studying 61,000 Microsoft employees, documented that cross-group collaboration dropped 25% under remote work, with networks becoming more static and siloed. The weak ties that used to carry information across the org went quiet. Leading remote engineers on top of that fragmented network is materially harder than leading the same people would have been pre-pandemic.
One important caveat. The proximity tax is not a case for return-to-office mandates. Bloom, Han, and Liang's June 2024 Nature paper studied 1,600 Trip.com employees and found two-day hybrid produced zero performance hit alongside a 33% attrition reduction. The data is clean. The problem is not where people work. The problem is what the manager role assumes it does not have to do.
Why the inherited model is obsolete
The foundational team-effectiveness research was conducted under a co-located default. Richard Hackman and Ruth Wageman's six conditions (real team, compelling direction, enabling structure, supportive context, expert coaching, adequate resources) carried up to 80% of the variance in team effectiveness in their studies. The six were not wrong. They were defined for teams that shared a building, and several of them depended on proximity to operate without ever saying so.
The same caveat applies to Google's Project Aristotle. The published re:Work materials identified psychological safety as the top of five dynamics across 180 teams. Psychological safety is a real and replicated construct. The instrument Aristotle used to study it, however, was a co-located instrument operating on co-located teams. The behavioral cues that produce safety, the recovery from interruption, the ability to read the room, the fast course-correction after a sharp word, all rely on physical presence in ways the original write-up did not have to name. The mechanism is real. The transfer to hybrid is non-trivial.
The inherited management curriculum, to the extent there is one, is even thinner. Gallup's 2024 research found only 44% of managers globally have received any formal training. Most managers became managers because they were good individual contributors. The promotion arrived without the training, and the training that does exist still teaches the co-located version of the job. Why new managers fail is, in the aggregate, a story of being handed a role whose defaults stopped working years before they were promoted.
The middle is where the system pressure shows up first. UKG and Gallup's 2024 work found 71% of middle managers overwhelmed, stressed, or burned out, with 76% of hybrid managers in that category and only 22% feeling their organization cares. The role is structurally untenable at scale. The fix is not more grit. It is a new mandate.
The chemistry-engineer mandate
The mandate has three working parts. None of them is novel in isolation. The novelty is making them the explicit work of management, rather than the implicit byproduct of being in the room.
1. Instrument signals
The chemistry-engineer manager runs lightweight, repeated diagnostics on the team rather than waiting for the quarterly engagement survey. Atlassian's Team Health Monitor is a quarterly eight-attribute self-assessment (shared understanding, value and metrics, suitable people, full-time owner, balanced team, proof of concept, one-pager, and velocity) that a team scores together. It is a coordination check, not a feelings check. The act of scoring together is most of the value.
Lara Hogan's BICEPS framework gives the manager a checklist to scan against in any moment of friction: Belonging, Improvement, Choice, Equality, Predictability, Significance. Most team upset traces back to a violation of one of the six. The framework turns a vague "something feels off" into a specific hypothesis about which need got bruised.
DORA's 2024 State of DevOps report identifies psychological safety and clear responsibilities as the strongest predictors of software-delivery performance, with unstable organizational priorities causing large productivity drops that the report describes as "highly resistant to mitigation." For an engineering manager, the DORA capabilities offer an empirical team-diagnostic spine that maps directly to delivery outcomes.
Signal does not require surveillance. Aggregate by default, opt-in, strengths-based, with a wall between coaching and reporting. A better engineering manager learns to read the team through behavior and structure, not through individual sentiment.
2. Design rituals that produce trust
Trust does not appear because people are well-intentioned. It is produced by repeated, low-stakes, intentionally vulnerable exposure to each other. The chemistry engineer designs that exposure on a cadence.
The cleanest hybrid protocol is Amy Edmondson and Mark Mortensen's five-step manager process: define the team's specific psychological-safety problem with the team itself, reframe disclosure norms to make small admissions normal, have the leader share their own constraints first, start small and iterate, then expand. The order matters. Leader vulnerability before team vulnerability. Small disclosures before large ones.
Patrick Lencioni's pyramid is older but still load-bearing. Trust enables productive conflict, which enables commitment, which enables accountability, which produces results. Skip the bottom layer and the rest of the stack does not hold. Lencioni's contribution is the ordering. You cannot install accountability into a team that does not have trust.
GitLab's all-remote handbook shows what a deliberate ritual layer looks like at scale. Mandatory coffee chats for new hires. A 2,700-page async substrate that lets the team operate without synchronous bottlenecks. In-person off-sites used surgically as chemistry events rather than work events. The discipline is not that everything is remote. The discipline is that the chemistry production is on the schedule, not in the hallway. Building trust on a remote team requires that the rituals be drawn into the calendar with the same seriousness as a sprint plan.
3. Schedule deliberate interdependence
The third pillar is structural. The chemistry engineer designs the team itself so that people who need to know each other actually have to work together.
Andy Grove's High Output Management introduced Task-Relevant Maturity (TRM) as the basis for varying 1:1 cadence by the report's experience with the specific task at hand. A senior engineer on familiar work gets monthly check-ins. The same engineer on an unfamiliar domain gets weekly. The cadence is a tool, not a courtesy.
Will Larson's Elegant Puzzle treats team chemistry as structural. Stable rosters. Six to eight reports per manager. Four states of a team: falling behind, treading water, repaying debt, innovating. Chemistry is engineered by sizing and load-shedding, not by morale events. A team in "falling behind" cannot be unstuck with a happy hour. It can be unstuck by changing the structure.
Camille Fournier's Manager's Path describes 1:1s as "oil changes" rather than emergency-response, and frames the manager's job as actively authoring the subculture of the team. The deliberate work is context-sharing across reports who would not naturally exchange context themselves. The org chart does not produce information flow. The manager does, through scheduling.
Where the metaphor leaks
Three counter-arguments deserve a hearing.
The manager-centric frame can overstate causal weight. The Gallup 70% number is comparative variance, not absolute causality. The remaining 30% includes compensation, role fit, organizational policy, and macro labor conditions. Critics have noted that manager-led interventions hit a ceiling without org-level change. The chemistry-engineer mandate has to be paired with org-level support, or the manager becomes the human shield for systemic problems. Gallup's own 2024 data on training drives this point home. 56% of managers globally have received no formal training. The role cannot deliver what the org has not invested in.
Chemistry may not be engineerable, and the metaphor can leak. Dennis Adsit and others have argued that the psych-safety-as-causal framing is over-claimed, and that several of the most-cited Aristotle-derived percentages do not appear in Google's published re:Work materials and may be fabricated downstream. The risk of taking the engineering metaphor too literally is that "chemistry instrumentation" tips into productivity theater, with managers performing data fluency while the team experiences surveillance. The mitigation is the same as for any team-data system: aggregate, opt-in, strengths-based, with a clean wall between coaching and reporting.
Hybrid is not the problem, RTO mandates are. Bloom's Trip.com data shows two-day hybrid is performance-neutral and attrition-reducing. Gartner's research finds no productivity gain from RTO mandates, with up to 80% of organizations losing top performers when mandates are imposed. The chemistry-engineer mandate should not be quoted to defend "just bring them back to the office." The work of management changed because the defaults changed. Reverting the location does not reinstall the defaults. It just adds a commute to a system that still needs intentional design.
What the new role actually looks like
The chemistry-engineer manager runs a stack that is concrete and small. A quarterly Atlassian Health Monitor. A BICEPS scan in any 1:1 where something feels off. A standing Edmondson-style disclosure ritual. A 1:1 cadence varied by Grove's TRM. A roster sized at six to eight reports, kept stable. A team operating model that names the four states and acts on them. A clean separation between aggregate team signal and private coaching. Off-sites used surgically for chemistry, not for work.
QuestWorks is one piece of that stack. The chemistry-engineer mandate calls for a deliberate ritual layer that produces behavioral signal without requiring the manager to be in the room. QuestWorks runs 25-minute AI-facilitated sessions for teams of two to five on its own cinematic, voice-controlled platform. The system observes the same coordination signals Anita Woolley's collective-intelligence research identifies as load-bearing: turn-taking, recovery after friction, network breadth. Outputs flow into a weekly team health report visible to leaders and a player-level QuestDash that is strengths-based and individual. Nine HeroTypes are public. Coaching with HeroGPT, which runs via Slack, is private. Slack and Microsoft Teams handle install, invites, leaderboards, and admin commands while the simulation runs on the QuestWorks platform. Participation is voluntary and not tied to performance reviews. Pricing is $14 per user per month for the first 50 companies in the Founder's Circle, $20 per user per month afterward, with a 10-day trial.
The category framing is straightforward. Team intelligence, powered by play. The mandate is to treat the team as a designable system. The manager is its engineer.
