A typical knowledge worker now spends a large share of the week in meetings, increasingly accompanied by AI that summarizes, recaps, and follows meetings on their behalf. The original promise was straightforward: fewer meetings, less drudgery, and more deep work. The evidence now suggests a more complex reality: meeting load can still rise, and attention can fragment faster than culture evolves.

This brief synthesizes meeting science, human-AI collaboration research, and enterprise diagnostics. The core finding is not “fewer meetings.” It is better meeting design plus better AI-enabled team habits. Teams that pair AI with strong norms convert individual productivity gains into collective outcomes; teams that do not simply accelerate existing calendar debt.

The TL;DR

  • Meetings are essential infrastructure for decisions, alignment, mentoring, and network formation. The issue is uneven quality, not meetings as a category [1] [16].
  • AI is increasing both meeting intensity and in-meeting multitasking in many contexts; this is not inherently negative, but it is design-sensitive [2] [3] [10].
  • Productive multitasking and distracted multitasking are different phenomena; dashboards should diagnose structure, not moralize behavior.
  • Individual AI adoption is ahead of team AI adoption; recap/follow/meeting hygiene are where team-level gains are unlocked [4] [20].
  • The practical agenda is to shrink bad meetings while protecting good ones — especially 1:1s, small decision forums, and cross-boundary collaboration contexts.
Part 1

Meetings are valuable — unevenly

Meeting research is unambiguous: well-designed meetings are fundamental to organizational performance, but poorly designed recurring meetings generate disproportionate cost [1] [16]. The “Lake Wobegon effect” in meetings persists — organizers tend to rate their meetings above average while attendees do not.

Meeting quality spectrum A spectrum from high-value small meetings to high-drift recurring meetings. The Meeting Quality Spectrum 1:1s and small decision meetings Working sessions (3-8) Cross-team syncs (5-12) All-hands and broadcasts (20+) Standing recurring meetings Lower drift risk → Higher drift risk when purpose, cadence, and attendee list are not actively managed.
Figure 1 — Meeting quality is mostly about design fit: right size, right duration, right cadence.
Meeting typePrimary valuePrimary risk when unmanaged
1:1s and small decision meetingsMentorship, decisions, trustUnder-scheduling can reduce alignment
Working sessions (3-8)Co-creation and problem-solvingDrift into status updates
Cross-team syncsBoundary coordinationPersist after objective is gone
Large broadcastsStrategic context and visibilityLow participation density, passive attendance
Recurring meetingsCadence and accountabilityHighest structural drift risk
The science does not say “fewer meetings.” It says right meeting, right size, right cadence.
Part 2

The multitasking taxonomy

Multitasking in meetings is not a single behavior with a single interpretation. Classic work identifies productive interleaving (notes, lookups, action capture) and distracted disengagement, both of which rise under different structural conditions [2] [3] [21].

Productive versus distracted multitasking Two-column taxonomy of productive and distracted multitasking patterns. The Multitasking Taxonomy Productive multitasking Distracted multitasking • Capturing action items in real time • Looking up data needed for the decision • Drafting follow-ups while context is fresh • Recap-assisted memory reinforcement • Follow-not-attend catch-up in focus blocks • Split-screen disengagement in low-value meetings • Chronic dual-attendance in conflicts • Large, long, recurring sessions with no agenda • Reactive meetings collapsing focus windows • No action-owner accountability after meetings Interpretation should be structural: high multitasking rates across a series often indicate meeting design debt, not individual failure.
Figure 2 — Productive and distracted multitasking are analytically distinct and should be diagnosed differently.
AI does not fix meeting culture. It amplifies the culture already present.
Part 3

What AI is actually doing to meetings

Recent research converges on a common pattern: AI speeds task execution but often increases total work throughput and parallelism [13] [14] [17]. In calendars, this frequently appears as more meetings, more overlap, and more recap-mediated catch-up behavior.

Signals of AI impact on meetings Five headline statistics presented as cards. Signals in the field +252% Teams meeting time vs. Feb 2020 62/mo Average meetings per worker 40-60 minutes/day individual AI savings 8h/mo Meeting content summarized async 37% Sustained users report attending fewer meetings Interpretation: growth and reduction signals can both be true at once. They usually describe different sub-populations and different norm maturity levels. AI deployment alone does not reduce meetings. Team norms determine whether saved time becomes better collaboration or more calendar load. Sources: Microsoft WorkLab/WTI, Atlassian State of Teams, New Future of Work synthesis.
Figure 3 — Individual gains are real; collective gains depend on team operating norms.
Part 4

The meeting benefits worth protecting

Calls to “reduce meetings” often cut the wrong layer. Three research streams show why: weak ties support mobility [6] [15], remote work can harden network silos [7] [22], and informal communication supports leadership and satisfaction [8] [18]. Good meeting policy protects those mechanisms while reducing structural waste.

Three benefits worth protecting Triad diagram showing weak ties, network bridging, and informal communication. What meetings protect when designed well Weak ties Career mobility and opportunity depend on cross-network contact. Network bridging Cross-boundary collaboration is easier in well-run sync contexts. Informal exchange Leadership trust and social glue often emerge around meetings. Protect these mechanisms while shrinking long, large, low-contribution recurring meetings.
Figure 4 — Meeting reduction without design precision can unintentionally weaken networks, leadership visibility, and career mobility pathways.
The meetings worth keeping build people. The meetings worth shrinking drain attention.
Part 5

Four practices to embrace

Individual AI adoption improves personal throughput. Team AI adoption improves collaborative outcomes. The following four practices are evidence-aligned and operationally simple.

Four practices stack Four stacked cards representing recap, follow-not-attend, focus catch-up, and meeting hygiene. Four practices for AI-era meeting culture 1. Use recap routinely across all meeting types 2. Accept one meeting, follow the conflicting one 3. Use focus blocks for recap catch-up and follow-through 4. Meeting hygiene: right-size, right-length, recurring audit, notice discipline
Figure 5 — These practices are designed to reduce bad meetings while preserving high-value collaboration contexts.

Practice 1 · Use recap routinely

Recap is the lowest-friction gain. It supports note quality, decision memory, and asynchronous catch-up. It is valuable not only for large meetings but also for 1:1s where presence and eye contact matter.

Copilot recap adoption by anonymized function Bar chart with anonymized functions A to E and adoption percentages. Copilot Recap Adoption by Function — Org B (anonymized) Regular recap usage (% of users) 02468 Function A7.0% Function B4.0% Function C3.0% Function D2.0% Function E2.0% Overall Org B baseline: 5.3% regular usage. Technology is deployed; behavior is uneven.
Figure 6 — Function labels are intentionally anonymized. Core signal retained: low baseline adoption and material variance across functions.

Practice 2 · Follow the conflict

When meetings conflict, “accept one and follow one” outperforms dual-attendance. This is especially effective in large meetings with low individual contribution density.

Practice 3 · Use focus time for catch-up

Batch recap processing in protected focus blocks. Fragmented recap checks throughout the day increase switch costs and residual cognitive load [21].

Practice 4 · Meeting hygiene

Four hygiene levers consistently outperform generic “reduce meetings” mandates: right-size attendee lists, default to shorter durations, audit recurring series quarterly, and enforce notice discipline.

Before during after scaffold Three-column checklist for before, during, and after meeting behaviors. Operational scaffold: Before · During · After Before During After • Right-size attendee list• 25/50 minute defaults• Clear agenda shared early• 24+ hour notice norm• Ask: should this be async? • Assign facilitator• Use AI for note/action capture• Keep contribution focus clear• Start/end on time• Resolve ownership live • Share recap within 24h• One owner + one deadline• Send recap to followers• Quarterly recurring audit• Track completion outcomes
Figure 7 — A simple before/during/after scaffold helps teams operationalize recap, follow, and hygiene practices consistently.
Part 6

Leadership take-aways

1) Aim at fewer bad meetings, not fewer meetings.

2) Treat individual AI adoption as necessary but insufficient.

3) Make recap adoption a named, measurable objective.

4) Audit variation by function and region, then intervene locally.

5) Empower managers first; they are leverage multipliers.

6) Pair license investment with norms investment.

7) Use structural levers: defaults, meeting-free windows, and organizer targeting.

Norms versus no norms trajectories Two lines showing diverging outcomes after AI deployment with and without norms. Two trajectories after AI deployment Low High Q1 Q2 Q3 Q4 AI + norms → better meetings, lower load AI without norms → meeting debt increases Collective productivity Calendar burden
Figure 8 — The divergence appears after deployment: norms determine whether AI time savings become collective gains or additional meeting load.
Measurement

What to measure: six signals worth tracking

  1. Meeting hours per week — baseline load signal.
  2. Large and long meeting share (9+ attendees and 60+ minutes) — strongest structural risk indicator.
  3. Multitasking rate — population-level quality signal, not an individual judgement.
  4. Available focus hours — whether deep work remains possible.
  5. After-hours collaboration — sustainability and burnout risk proxy.
  6. Late join / late end frequency — operational discipline indicator.

A coda. The meetings worth keeping build people: the 1:1s where mentorship happens, the small sessions where decisions are made, and the cross-team conversations where weak ties form. The meetings worth shrinking drain attention: long recurring broadcasts, stale syncs, and reactive fire-drills. AI will amplify whichever system teams choose to build.

References

  1. Rogelberg, S. G. (2019). The Surprising Science of Meetings. Oxford University Press.
  2. Iqbal, S. T., Grudin, J., & Horvitz, E. (2011). Peripheral computing during presentations. CHI.
  3. Cao, H. et al. (2021). Large-scale analysis of multitasking during remote meetings. CHI.
  4. Butler, J. et al. (Eds.). (2025). Microsoft New Future of Work Report 2025.
  5. Microsoft & LinkedIn. (2024). Work Trend Index Annual Report.
  6. Rajkumar, K. et al. (2022). A causal test of the strength of weak ties. Science.
  7. Yang, L. et al. (2022). Effects of remote work on collaboration. Nature Human Behaviour.
  8. Lutjens, M., & Felfe, J. (2025). Informal communication and job satisfaction in hybrid work.
  9. Microsoft. (2025). Work Trend Index Annual Report.
  10. Microsoft WorkLab. (2022). Too many meetings? Here's how AI could change that.
  11. Microsoft WorkLab. (2024). AI Data Drop: The 11-by-11 tipping point.
  12. Atlassian. (2025). State of Teams 2025.
  13. Ranganathan, A., & Ye, A. (2026). AI doesn't reduce work — it intensifies it. HBR.
  14. Brynjolfsson, E., Li, D., & Raymond, L. (2025). Generative AI at work. QJE.
  15. Granovetter, M. (1973). The strength of weak ties. AJS.
  16. Allen, J. A., Lehmann-Willenbrock, N., & Rogelberg, S. G. (Eds.). (2015). The Cambridge Handbook of Meeting Science.
  17. Dell'Acqua, F. et al. (2025). Navigating the jagged technological frontier. Organization Science.
  18. Saatci, B. et al. (2020). Reconfiguring hybrid meetings. CSCW.
  19. Amershi, S. et al. (2019). Guidelines for Human-AI Interaction. CHI.
  20. Asthana, S. et al. (2024). LLM-powered meeting recap system. PACM HCI / CSCW.
  21. Mark, G., Gudith, D., & Klocke, U. (2008). The cost of interrupted work. CHI.
  22. Olson, G. M., & Olson, J. S. (2000). Distance matters. HCI.

Org A = anonymized European mobility/IT enterprise diagnostic. Org B = anonymized global luxury brand diagnostic. Metrics shown in aggregate only.