AI Meeting Assistants Split Over Bots, Consent and Agents in 2026

A crowded category settles on a short list

In 2026, nearly every productivity publication and software vendor has published its own ranking of AI meeting assistants. The lists vary in length from a handful of tools to more than two dozen, yet they mostly name the same products. Zapier's April roundup put Fireflies, Granola, Avoma, tl;dv, Krisp, Otter, Fellow, MeetGeek and Fathom alongside Calendly's note-taker.11 TechRepublic's September list kept six tools: Fathom as its overall pick, Fireflies.ai for teams with heavy meeting loads, Otter.ai for live transcription, Granola for capture without a bot, Fellow for structured team meetings and Notta for bilingual calls.4 A comparison published in October by a rival vendor arrived at much the same group, recommending Fathom for general use, Fireflies for sales teams and Granola for solo users who don't want a bot in the call.1

The agreement is the story here. When independent reviewers, competitors and affiliate sites keep landing on the same four or five products, the notetaker market looks less like an open race and more like a maturing category. Its leaders are known, and they now compete mostly on price, privacy defaults and integrations.

The prices themselves show how commoditized the core feature has become. Reviewers list Otter's paid tier from $8.33 per user per month on annual billing and Fireflies Pro from $10 per seat.84 Granola's Business plan is listed at $14 per user.4 Fathom's free plan is widely cited as the most generous, with unlimited recordings for individuals.919 If transcription and summaries cost less than a streaming subscription, or nothing at all, vendors have to find another way to stand out.

The bot problem became the dividing line

The most consistent theme across the 2026 coverage is the divide between tools that send a visible bot into the meeting and tools that record locally on the user's device. One review describes the market as split by "the bot problem," and names Granola, Jamie and Krisp as the tools that avoid a bot completely.19 Another sets out the trade-offs. Bot-based tools receive cleaner audio and name speakers more reliably. Botless tools stay out of sight of clients but depend on the user's microphone and are weaker at telling speakers apart.6

The incumbents have responded by offering both options. Otter, Fireflies and Fathom now sell botless desktop apps alongside their meeting bots, though Fathom's version is still described as a Mac beta.5 One reviewer notes that Granola was designed from the start around local capture, while Otter and Fireflies added bot-free modes later.2 Accounts of Granola's reach differ. One September hands-on review called it Mac-only.6 Other recent coverage says it now runs on Windows, and one review also lists iPhone and Android.15 The difference probably reflects how quickly the product is changing rather than a real dispute. Still, it shows that many "2026" lists are snapshots that go stale within months.

The Otter litigation changed the privacy question

The shift toward invisible capture did not happen in isolation. In August 2025, a proposed class action accused Otter.ai of recording meeting participants without the consent of everyone present and of using those conversations to train its speech models.33 The lead plaintiff said he had no Otter account and had joined a Zoom call where someone else was running the Otter Notetaker.38 Four suits were later consolidated in the Northern District of California as In re Otter.AI Privacy Litigation.39 Otter's defense is that consent belongs to the account holder, and it argued that no interception occurred.35

A court opinion dated August 13, 2026 rejected parts of Otter's motion to dismiss. It found that the plaintiffs' alleged injury closely resembles the common-law privacy tort of intrusion upon seclusion. It also rejected the idea that a notetaker appearing as a meeting participant proves the other attendees knew they were being recorded and stored, at least for one Washington plaintiff. The court did agree with Otter that another plaintiff's claim to a reasonable expectation of privacy was weak.40 Reviewers have taken notice. One October guide tells readers to consider "the ongoing class-action context" before choosing Otter.1 An April test of 14 tools flagged the lawsuit as something buyers should know about.3

The lawsuit has an awkward consequence for the bot-free pitch. Some reviewers praise botless tools precisely because clients "never know you're recording."6 Yet the central allegation in the Otter case is recording without everyone's consent. A European competitor argued that removing the bot moves the problem rather than solving it, because the recording still happens and only its visibility goes away.37 That argument is persuasive. A tool sold on being undetectable fits poorly with all-party consent laws, and the safer setup combines a visible bot, a spoken announcement and contractual limits on training. Data location has also become a selling point. tl;dv is recommended for storing data in the EU, and one review lists whether each vendor trains models on customer content.16

Transcription is cheap now. The value is in what happens next.

The second clear trend is that vendors are moving from note-taking to taking action. Tana argues that 2026 tools fall into two groups: notetakers that record action items after the call, and "agentic" platforms whose AI files tickets and drafts documents during the meeting.14 Simular, which tested eight tools over six weeks, concluded that most products stop at transcripts and summaries, and that the more important question is what happens after the notes are written.15 Hedy is marketing live, private coaching during conversations.5 Platform owners are bundling the feature into products companies already pay for: Microsoft Copilot, Zoom AI Companion and Gemini in Google Meet all appear in recent comparisons as defaults for organizations standardized on those suites.78

Many of these claims come from vendors promoting their own products, which is a reason for caution. The direction is still clear. Once transcription costs almost nothing and comes built into Zoom, Teams and Meet, independent vendors can only defend their prices by tying meetings to CRMs, project boards and email follow-ups. This is why Fireflies is repeatedly singled out for its CRM integrations.23

Adoption is broad but shallow

Survey data puts the spread of these tools in context. Gallup reported in May 2026 that 52% of U.S. employees use AI in their role. Only 30% use it a few times a week or more, and 15% use it daily.30 Forty-seven percent said their organization has integrated AI tools, up from 41% the previous quarter.25 Productivity gains are concentrated among people who use AI for many different tasks. Forty-five percent of employees using AI for one or two purposes reported a positive effect, compared with 90% of those using it for seven or more.22

The broader picture is less clear-cut. SHRM found that workers estimate AI saves them about six hours a week.23 The same SHRM data also shows about four hours a week spent correcting AI output.26 Only 14% of employees at companies using AI strongly agree that it has changed how work gets done.30 A 2026 NBER survey of nearly 6,000 executives found that 89% saw no measurable effect on labor productivity over three years.29 Meeting notetakers match this pattern well. They save time on a specific task that is easy to measure, but they rarely change how an organization works unless the summaries feed into systems that trigger action.

The labor question falls on administrative work

The labor effects are most direct in the jobs these tools automate. The Bureau of Labor Statistics projects declines through 2034 of 5.8% for legal secretaries and administrative assistants, 4.9% for medical transcriptionists and 1.6% for general secretaries and administrative assistants, and it cites AI-driven efficiency as a factor.46 Associated Press reporting noted that the number of secretaries and administrative assistants fell from 3.5 million in 2004 to 2.1 million in 2024. A Brookings report warned that clerical workers, about 86% of them women, may lack the "adaptive capacity" to change careers. One Vanderbilt executive assistant told the AP that she no longer takes notes in meetings and finishes work that used to take hours in under five minutes.45

Across the economy, the evidence is mixed. Yale's Budget Lab finds no clear effect of AI on overall employment. Stanford researchers estimate that employment for 22- to 25-year-olds in highly exposed occupations is about 19% below where it would otherwise be.

AI meeting assistants are not a primary cause of job losses. They are a clear example of how displacement tends to happen in practice: one routine task disappears at a time, and the people who keep their jobs move toward judgment and coordination. The consent litigation will help decide how these tools spread. The labor data shows who absorbs the cost of the time they save.