Three out of four professionals (75%) now use an AI note-taker in their work meetings, roughly double the share in 2023 (Fellow.ai, State of AI Meeting Notetakers 2025). The tools they rely on have reached real scale: Otter.ai crossed 1 billion cumulative meetings and 35 million users at 100 million dollars in ARR (Otter.ai, 2025), Microsoft’s Copilot family passed 150 million monthly active users (Microsoft, FY2026 Q1 earnings), and Zoom’s AI Companion paid users grew 184% year over year into early 2026 (Zoom, Q1 FY2027 results). The money is following the usage, with the AI meeting assistants market valued between 1.42 and 4.31 billion dollars for 2026 depending on scope (Precedence Research; Grand View Research, 2026). This analysis consolidates data from Fellow.ai, Microsoft, Zoom, Otter.ai, and 15 other primary sources into one reference on AI meeting assistants by the numbers.
TL;DR
- 75% of professionals use an AI note-taker in work meetings, up from about 25% in 2023 (Fellow.ai, State of AI Meeting Notetakers 2025).
- The AI meeting assistants market is valued at 1.42 billion dollars (Precedence Research) or 4.31 billion dollars (Grand View Research) for 2026, reflecting different scope definitions (2026).
- Grand View Research forecasts 21.48 billion dollars by 2033 at a 25.8% CAGR; Precedence sees 6.28 billion by 2035 at 18% (2026).
- Otter.ai reached 1 billion+ cumulative meetings, 35 million+ users, and 100 million dollars ARR in 2025 (Otter.ai, 2025).
- Fireflies.ai reported 20 million+ users across 500,000+ organizations and a 1 billion dollar valuation (Fireflies.ai, June 2025).
- Microsoft’s Copilot family passed 150 million monthly active users, with 20 million+ paid Microsoft 365 Copilot seats by April 2026 (Microsoft, FY2026 earnings).
- Google Meet’s “Take Notes for Me” was used by 110 million+ attendees in a single month, up 8.5x year over year (Google Workspace, 2026).
- Employees are interrupted every 2 minutes, about 275 times a day, and 57% of meetings are ad hoc (Microsoft, Work Trend Index 2025).
- 62% of AI meeting-tool users save roughly 4 hours per week, and action-item completion rises from 50 to 60% up to 85 to 95% (Laxis, State of Meeting Note-Taking 2026).
- Leading transcription engines hit 5.6% word error rate on English overall, but multi-speaker clinical conversations can exceed 50% (AssemblyAI 2026; systematic review, PMC 2025).
- 84% of users change what they say when an AI note-taker is present, and 47% saw a tool capture something unintended (Fellow.ai, 2025).
- Otter.ai faces a consolidated class action in the Northern District of California bundling four suits filed in August and September 2025 (NPR, 2025).
1. Market Size and Growth
The headline number depends almost entirely on where each firm draws the boundary. Precedence Research values the category at 1.42 billion dollars for 2026, while Grand View Research puts it at 4.31 billion dollars for the same year, a roughly 3x spread driven by whether the definition counts only dedicated notetakers or the wider assistant layer inside conferencing suites. The direction is not in dispute: every firm models double-digit growth, with CAGR estimates clustering between 18% and 26%. A narrower “AI note-taking” segment tracked by Precedence Research sits at 623.5 million dollars for 2025, a reminder that pure notetaking is a subset of the larger assistant market. Cross-referenced against Grand View Research, Market Research Future’s 25.62% CAGR to 2035 lands inside the same growth band.
| Metric | Value | Source |
|---|---|---|
| AI meeting assistants market, 2025 (narrow scope) | 1.20 billion USD | Precedence Research (2026) |
| AI meeting assistants market, 2025 (broad scope) | 3.47 billion USD | Grand View Research (2026) |
| AI meeting assistants market, 2026 | 1.42 billion (Precedence) / 4.31 billion (Grand View) USD | Precedence Research; Grand View Research (2026) |
| Forecast market size, 2033 | 21.48 billion USD | Grand View Research (2026) |
| Forecast market size, 2035 | 6.28 billion USD | Precedence Research (2026) |
| Projected CAGR | 18.0% (Precedence) to 25.8% (Grand View) | Precedence Research; Grand View Research (2026) |
Outlier note: the 3x gap between the low and high 2026 estimates is a scope artifact, not a disagreement about growth. Cloud-based deployment held about 75% of the market and the software segment over 70% in 2025 breakdowns, and North America led with roughly a third of revenue.
2. Adoption: Who Uses AI Notetakers
Adoption is high on average but strikingly uneven by company size. The 75% adoption headline masks a curve that peaks at 74% for firms of 201 to 1,000 employees and falls to 43% at organizations over 5,000 staff (meetingstack, 2026). The bottleneck at the top is not the tooling but procurement, legal review, and shadow-app risk. Vertical splits tell the same story: developer- and sales-heavy sectors are saturated while regulated industries lag, with government at 11% against SaaS at 72%. meetingstack notes its model relies on public signals (review volume, job posts, integrations) and therefore likely understates stealth adoption.
| Metric | Value | Source |
|---|---|---|
| Professionals using an AI note-taker | 75% | Fellow.ai, State of AI Meeting Notetakers 2025 |
| Adoption, 1-10 employees | 28% | meetingstack, AI Notetaker Adoption (2026) |
| Adoption, 201-1,000 employees (peak) | 74% | meetingstack, AI Notetaker Adoption (2026) |
| Adoption, 5,000+ employees | 43% | meetingstack, AI Notetaker Adoption (2026) |
| Adoption, SaaS / tech vertical | 72% | meetingstack, AI Notetaker Adoption (2026) |
| Adoption, government vertical | 11% | meetingstack, AI Notetaker Adoption (2026) |
| Overall adoption, 2023 to 2026 | 25% to 67% | meetingstack, AI Notetaker Adoption (2026) |
| Most-used tools among users | Fathom 26% / Otter.ai 22% / Fireflies.ai 16% | meetingstack, AI Notetaker Adoption (2026) |
Outlier note: mid-market firms out-adopt the Fortune 500 here, inverting the usual enterprise-first pattern for new software.
3. Platform Scale: Zoom, Microsoft, Google, Otter, Fireflies, Gong
No single vendor dominates disclosed usage, which is why the category still reads as a land grab. Otter.ai passed 1 billion cumulative meetings and 35 million users while Fireflies.ai reported 20 million users across 500,000 organizations (Otter.ai, 2025; Fireflies.ai, June 2025). The platform incumbents move a different lever: Zoom’s AI Companion paid users grew 184% year over year (Zoom, Q1 FY2027 8-K), Google Meet’s “Take Notes for Me” reached 110 million+ attendees in a month, and Microsoft’s Copilot family hit 150 million monthly active users. For readers comparing the underlying conferencing platforms, our video conferencing statistics roundup covers Zoom and Teams share in detail.
| Metric | Value | Source |
|---|---|---|
| Zoom AI Companion paid MAU growth | +184% year over year | Zoom, Q1 FY2027 results (May 2026) |
| Zoom “My Notes” licensed users | 1.5 million in 4 months | Zoom, Q1 FY2027 results (May 2026) |
| Microsoft Copilot family MAU | 150 million | Microsoft, FY2026 Q1 earnings |
| Microsoft 365 Copilot paid seats | 20 million+ (by April 2026) | Microsoft, reported by TechCrunch (2026) |
| Google Meet “Take Notes for Me” attendees | 110 million+/month, up 8.5x YoY | Google Workspace (2026) |
| Otter.ai cumulative meetings / users / ARR | 1 billion+ / 35 million+ / 100 million USD | Otter.ai (2025) |
| Fireflies.ai users / organizations / valuation | 20 million+ / 500,000+ / 1 billion USD | Fireflies.ai (June 2025) |
| Gong ARR run rate | 500 million USD, +55% YoY, 5,000+ companies | Gong, press release (2025) |
Outlier note: Microsoft’s 150 million “Copilot family” figure spans many products; its dedicated Microsoft 365 Copilot paid base was about 20 million seats against a 450 million commercial installed base by early 2026 (Microsoft; TechCrunch, 2026).
4. The Meeting Problem AI Is Built to Solve
Adoption makes sense once you count the meetings. Microsoft’s 2025 Work Trend Index found employees are interrupted every 2 minutes, roughly 275 times a day, and spend 57% of their time communicating versus 43% creating (Microsoft, Breaking Down the Infinite Workday, 2025). More than half of meetings are unplanned, half cluster into the same peak-productivity windows, and after-hours meetings keep climbing. That fragmentation is exactly the workflow an assistant absorbs by capturing, summarizing, and routing action items so humans do not have to. The distributed nature of modern teams compounds it, a pattern we track in our remote work statistics roundup.
| Metric | Value | Source |
|---|---|---|
| Time spent communicating vs creating | 57% / 43% | Microsoft, Work Trend Index 2025 |
| Daily interruptions per worker | 275 (one every 2 minutes) | Microsoft, Work Trend Index 2025 |
| Meetings in peak-productivity windows (9-11am, 1-3pm) | 50% | Microsoft, Work Trend Index 2025 |
| Ad hoc meetings (no calendar invite) | 57% | Microsoft, Work Trend Index 2025 |
| Meetings after 8pm, year-over-year change | +16% | Microsoft, Work Trend Index 2025 |
| Workers saying work feels chaotic/fragmented | 48% employees / 52% leaders | Microsoft, Work Trend Index 2025 |
| Cost of unproductive meetings, US (most recent widely cited) | up to 399 billion USD/year | Runn, Unproductive Meetings Statistics (2024) |
Outlier note: PowerPoint edits spike 122% in the final 10 minutes before a meeting, a signal of how much prep happens at the last second (Microsoft, Work Trend Index 2025). The 399 billion dollar cost figure derives from older 2022-era modeling and is flagged as the most recent widely cited estimate.
5. ROI and the Productivity Payoff
Time saved is the metric that keeps budgets flowing, and the numbers are consistent across independent sources. Laxis reports that 62% of AI meeting-tool users reclaim about 4 hours per week, roughly a month of work per person per year, while action-item completion climbs from a 50 to 60% baseline up to 85 to 95% (Laxis, State of Meeting Note-Taking 2026). Vendor-reported returns line up: Otter.ai cites a 10-to-1 ROI and one full-time-employee equivalent saved for every 20 users, and Read AI says its users attend 20% fewer meetings. The through-line is that the payoff comes less from faster note-typing than from fewer recap meetings once a searchable record exists.
| Metric | Value | Source |
|---|---|---|
| Users saving about 4 hours per week | 62% | Laxis, State of Meeting Note-Taking 2026 |
| Action-item completion, before vs after AI summaries | 50-60% up to 85-95% | Laxis, State of Meeting Note-Taking 2026 |
| Meeting-length reduction within 6 months of default capture | 20-40% | Laxis, State of Meeting Note-Taking 2026 |
| Otter.ai enterprise ROI ratio | 10-to-1 (1 FTE saved per 20 users) | Otter.ai (2025) |
| Otter.ai cumulative customer ROI generated | 1 billion USD+ | Otter.ai (2025) |
| Read AI users’ meeting reduction | 20% fewer meetings attended | Read AI (2026) |
Outlier note: Otter’s own ROI figures are vendor-reported; Laxis’s are survey-based. Both point the same direction but should be read as self-reported rather than audited.
6. Transcription Accuracy: the Technical Ceiling
Accuracy is where marketing collides with acoustics. Leading engines post about 5.6% word error rate on English overall and 1.52% on clean studio audio, but a live multi-speaker meeting is a different problem entirely (AssemblyAI, 2026). AssemblyAI’s own real-world table drops to 85 to 92% accuracy on video calls and 70 to 85% in noisy rooms, and a clinical systematic review found multi-speaker conversations can exceed 50% word error rate against roughly 8.7% for controlled dictation. That gap is why summaries built on top of shaky transcripts can quietly misattribute quotes or drop names. The same speech-recognition frontier shapes consumer dictation, including features in tools like VoxBooster’s own dictation and voice control; our speech-to-text statistics roundup goes deeper on the benchmarks.
| Metric | Value | Source |
|---|---|---|
| Leading model English mean word error rate | 5.6% (median 4.9%) | AssemblyAI, Universal-3 Pro (2026) |
| LibriSpeech Clean word error rate | 1.52% | AssemblyAI (2026) |
| Earnings21 (real business-call) word error rate | 8.80% | AssemblyAI (2026) |
| Video conference call transcription accuracy | 85-92% | AssemblyAI (2026) |
| Noisy-environment transcription accuracy | 70-85% | AssemblyAI (2026) |
| Controlled medical dictation word error rate | ~8.7% | Systematic review, PMC (2025) |
| Multi-speaker clinical conversation word error rate | exceeds 50% | Systematic review, PMC (2025) |
Outlier note: the 30x-plus spread between clean-audio WER (1.52%) and worst-case multi-speaker WER (50%+) explains why identical tools earn wildly different reviews depending on room and headset.
7. Privacy, Consent, and Legal Risk
The fastest-growing objection to AI notetakers is not accuracy but consent. 84% of users say they change what they say when an AI note-taker is present, 47% have seen a tool record or share something unintended, and 50% of non-users name privacy as their top reason to abstain (Fellow.ai, 2025). Those behavioral shifts are now backed by litigation: Otter.ai faces a consolidated class action in California federal court, where all-party consent law is central, and institutions are drawing lines, with Harvard IT advising against unapproved assistants. The chilling effect is real, and it explains the 2026 pivot toward “bot-free” capture that avoids adding a visible participant.
| Metric | Value | Source |
|---|---|---|
| Non-users citing privacy/security as top barrier | 50% | Fellow.ai, State of AI Meeting Notetakers 2025 |
| Users altering what they say with a notetaker present | 84% | Fellow.ai, State of AI Meeting Notetakers 2025 |
| Active users who saw unintended capture/sharing | 47% | Fellow.ai, State of AI Meeting Notetakers 2025 |
| Otter.ai suits consolidated in N.D. California | 4 (filed Aug-Sep 2025) | NPR, reporting court filings (2025) |
| Otter.ai motion-to-dismiss hearing | scheduled May 20, 2026 | N.D. Cal. docket, via Recording Law (2026) |
| Harvard IT restriction on unapproved assistants issued | Feb 11, 2025 | Harvard University IT (2025) |
Outlier note: California and Illinois are consent-law flashpoints; Fireflies.ai separately faces a December 2025 suit under Illinois’s biometric privacy law (BIPA). See NPR’s coverage and Harvard IT guidance.
Summary: AI Meeting Assistants by the Numbers
| Metric | Value | Source |
|---|---|---|
| Professionals using an AI note-taker | 75% | Fellow.ai, State of AI Meeting Notetakers 2025 |
| AI meeting assistants market, 2026 | 1.42 billion (Precedence) / 4.31 billion (Grand View) USD | Precedence Research; Grand View Research (2026) |
| Overall adoption, 2023 to 2026 | 25% to 67% | meetingstack, AI Notetaker Adoption (2026) |
| Otter.ai cumulative meetings / users / ARR | 1 billion+ / 35 million+ / 100 million USD | Otter.ai (2025) |
| Fireflies.ai users / organizations / valuation | 20 million+ / 500,000+ / 1 billion USD | Fireflies.ai (June 2025) |
| Microsoft Copilot family MAU | 150 million | Microsoft, FY2026 Q1 earnings |
| Zoom AI Companion paid MAU growth | +184% year over year | Zoom, Q1 FY2027 results (2026) |
| Zoom “My Notes” licensed users | 1.5 million in 4 months | Zoom, Q1 FY2027 results (2026) |
| Google Meet “Take Notes for Me” attendees | 110 million+/month, up 8.5x YoY | Google Workspace (2026) |
| Gong ARR run rate | 500 million USD, +55% YoY | Gong, press release (2025) |
| Daily interruptions per worker | 275 (one every 2 minutes) | Microsoft, Work Trend Index 2025 |
| Cost of unproductive meetings, US | up to 399 billion USD/year | Runn (2024) |
| Users saving about 4 hours per week | 62% | Laxis, State of Meeting Note-Taking 2026 |
| Action-item completion after AI summaries | up to 85-95% | Laxis, State of Meeting Note-Taking 2026 |
| Leading model English word error rate | 5.6% | AssemblyAI (2026) |
| Video conference call transcription accuracy | 85-92% | AssemblyAI (2026) |
| Multi-speaker clinical conversation WER | exceeds 50% | Systematic review, PMC (2025) |
| Otter.ai enterprise ROI ratio | 10-to-1 | Otter.ai (2025) |
Methodology and Sources
Data was gathered by aggregating primary reports, company disclosures, earnings filings, survey datasets, court filings, and peer-reviewed research published mainly in 2025 and 2026, with each figure traced back to its originating source rather than a secondary quote. Sources cited:
- Fellow.ai, State of AI Meeting Notetakers 2025 (link)
- meetingstack, AI Notetaker Adoption (2026) (link)
- Laxis, State of Meeting Note-Taking 2026 (link)
- Precedence Research, AI in Meeting Assistants Market and AI Note Taking Market (2026) (link)
- Grand View Research, AI Meeting Assistant Market Report, 2026-2033 (link)
- Market Research Future, AI Meeting Assistants Market (2035 outlook)
- Microsoft, Work Trend Index 2025: Breaking Down the Infinite Workday (link)
- Microsoft, FY2026 Q1 earnings (Copilot family MAU); Microsoft 365 Copilot paid seats reported by TechCrunch (2026)
- Zoom, Q1 FY2027 results (8-K, May 2026) and Q1 FY2026 results (link)
- Google Workspace, Take Notes for Me usage (2026) (link)
- Otter.ai, 2025 milestone press releases (link)
- Fireflies.ai, 1 billion dollar valuation announcement (June 2025) (link)
- Gong, ARR tops 500 million dollars press release (2025)
- Read AI, product usage claims (2026)
- Runn, Unproductive Meetings Statistics (2024) (link)
- AssemblyAI, How accurate is speech-to-text (2026) (link)
- Peer-reviewed systematic review of AI speech recognition for clinical documentation, PMC (2025) (link)
- NPR, reporting on the Otter.ai class action (2025) (link)
- Harvard University IT, AI Assistant Guidelines (2025) (link)
Data watch: Microsoft’s Work Trend Index publishes a major edition annually (next expected in 2026), Fellow.ai and Laxis both run recurring “State of” surveys likely to refresh in late 2026, Zoom and Microsoft report quarterly earnings, and Grand View Research and Precedence Research update their market models each year. Otter.ai’s California class action has a motion-to-dismiss hearing set for May 20, 2026, which may materially change the legal picture.
Last updated: July 5, 2026. We review and update this page quarterly as new data is published.