Fellow logo
Meeting Assistants · Audio Transcription

Fellow

Fellow is an AI meeting assistant that records, transcribes and summarises calls across Zoom, Teams, Google Meet and in-person rooms, then turns them into action items and CRM updates under admin-controlled privacy, retention and redaction rules.

Active GDPR compliant Free plan · Free trial Freemium API available 13+ Verified by Guidaio
Overview

What is Fellow?

Fellow is an AI meeting assistant published by Fellow Insights Inc. of Ottawa, founded in November 2017 by Aydin Mirzaee, Amin Mirzaee and Samuel Cormier-Iijima — the team behind Fluidware before SurveyMonkey acquired it. Shopify was the first customer. The product started as collaborative agendas and one-on-one notes, added AI notetaking in 2023, and was renamed from Fellow.app to Fellow.ai in 2025.

It captures meetings on Zoom, Microsoft Teams, Google Meet, Webex, phone and in person. Two capture modes run under the same governance framework: a bot that joins the call and asks to be admitted, announcing the recording, or botless capture through the native macOS and Windows desktop app, with speaker diarisation for conference rooms. Recordings, transcripts, summaries, decisions and action items land in a permissioned library. AskFellow, the built-in AI agent, answers questions across every meeting a user can access and drafts documents, memos, follow-up emails and CRM updates — Salesforce and HubSpot fields are suggested from what was actually said.

The governance layer is what sets Fellow apart in a crowded category. Administrators set recording policies per organisation, team or user, block capture by counterparty domain or by keyword in the meeting title, and turn on pre-meeting disclosure emails that are logged for review. Users pause and resume recording, and the pause itself is timestamped. Transcript redaction removes names, account numbers, NPI and MNPI from the transcript, the summary and the audio. Zero-Day Retention deletes source recordings and transcripts straight after processing while keeping the AI summary as institutional memory. Deletions are written to an exportable audit trail via the Super Admin API, with archiving to Global Relay, and information barriers keep business units apart.

That aim is deliberate: private equity firms, hedge funds, investment banks, RIAs and broker-dealers under SEC recordkeeping and FINRA supervision, alongside legal, health tech and IT teams. Fellow claims SOC 2 Type II — audited since 2020 — HIPAA with a BAA on request, and GDPR compliance, with AES-256 at rest and TLS 1.2+ in transit. Infrastructure runs on AWS in the Canada Central region. Customer data is never used to train AI models. A public REST API, an MCP server and a Claude connector complete the platform, next to more than forty named integrations.

What it does

  • Record and transcribe meetings with or without a bot, across Zoom, Microsoft Teams, Google Meet, Webex, phone and in-person rooms
  • Generate AI summaries, decisions and assignable action items the moment a meeting ends
  • Ask AskFellow questions across every meeting you can access, and have it draft documents, memos and follow-up emails
  • Push meeting insights into Salesforce and HubSpot as suggested field updates
  • Prepare ahead with collaborative agendas and pre-meeting briefs
  • Centralise every recording in a permissioned library with custom channels
  • Enforce retention policies, transcript redaction and exportable audit logs through the Super Admin API
Audience

When to use Fellow / When not to

A quick filter to help you decide if Fellow is the right fit.

When to use Fellow

  • Managers and executives who live in back-to-back meetings and need reliable recaps, decisions and action items without taking notes themselves
  • Regulated finance teams — private equity, hedge funds, investment banks, RIAs and broker-dealers working under SEC recordkeeping and FINRA supervision
  • Sales and customer success teams that want call insights pushed straight into Salesforce or HubSpot instead of retyped after the fact
  • IT, security and compliance owners who need org-wide governance over who may record, what may be recorded and how long anything is kept
  • Product, engineering, HR and legal teams whose discovery calls, interviews and reviews are worth searching later rather than forgetting

When not to use Fellow

  • Teams needing a genuinely usable free tier — the free plan allows five AI notes and five AI recordings per user for life, not per month
  • Anyone who needs a non-English interface: every page collected is English-only, and no localised UI is advertised
  • Clinicians looking for a medical scribe — Fellow is HIPAA-ready for business meetings, not a clinical documentation tool
  • Small teams wanting enterprise safeguards on a cheap plan: redaction, advanced recording permissions, SSO, HRIS sync and the HIPAA BAA all sit on the Enterprise tier
  • Developers who expect API access from day one — the API starts on the paid Team plan, and Enterprise pricing requires a sales conversation from ten seats up
Get started

How to use Fellow

A typical end-to-end flow, from setup to results.

  1. Create a free account from the sign-up page — no credit card is required
  2. Connect your calendar (Google Workspace, Office 365 or iCloud) so Fellow can see your meetings
  3. Sign in through Google or Office 365 SSO; Okta, OneLogin and custom OIDC are available on Enterprise
  4. Choose your capture mode: let the bot join and ask to be admitted, or install the macOS/Windows desktop app for botless recording
  5. Install the iOS or Android app for in-person conversations, and the Chrome extension if you work in the browser
  6. Set which meetings Fellow joins automatically, or invite it manually from your conferencing tool
  7. Prepare with a collaborative agenda or a pre-meeting brief before the call starts
  8. After the meeting, review the summary, decisions and action items, then share them through a permissioned channel
  9. Admins: configure recording policies, retention schedules, redaction rules and the AI toggle in Workspace Settings
  10. Developers: generate an API key from your workspace, revocable at any time, and connect through the REST API or the MCP server
Quick read

Pros & Cons

Pros

  • Recording governance far deeper than the category norm: policies by organisation, team, user, counterparty domain or keyword
  • Both bot and botless capture, sold as a per-meeting choice under one security framework
  • A clear no-training commitment on customer data, restated in three separate places on the site
  • A dated, public sub-processor list that names data locations and separates optional AI sub-processors
  • Application data hosted on AWS in the Canada Central region, documented explicitly rather than implied
  • SOC 2 Type II audited since 2020, HIPAA with a BAA on request, and annual third-party penetration tests
  • A mature product — founded 2017, two announced funding rounds, desktop, iOS, Android and browser clients

Cons

  • The free plan gives five AI notes and five AI recordings for life per user, not per month, so it is a demo rather than a workable tier
  • Seat caps are rigid: five on Free, twenty on Team, fifty on Business, with unlimited seats only on Enterprise
  • Transcript redaction, advanced recording permissions, SSO, HRIS sync and the HIPAA BAA are all Enterprise-only
  • API access starts on the paid Team plan, and Enterprise requires a sales conversation from ten seats upwards
  • The transcription language count contradicts itself across the site — 90+, 92 or 99 depending on the page — and no list of languages is published
  • The SOC 2 Type II report is shared only on request under a mutual NDA, and the Vanta trust centre renders nothing without JavaScript
  • No EU entity, no EU address and no named Article 27 representative; transfers rest on Standard Contractual Clauses
Pricing

Pricing & Plans

Fellow offers a permanent free plan at 0 USD, capped at five seats with five AI notes and five AI recordings per user for the lifetime of the account. The cheapest paid entry point is the Team plan at 7 USD per user per month billed annually, or 11 USD billed monthly. Business is 15 USD per user per month annually (23 USD monthly), and Enterprise is 25 USD per user per month annually from ten users, through sales. A 14-day free trial of the Fellow Pro plan is available with AI recording credits, after which the workspace reverts automatically to the free plan. All prices are quoted in US dollars.

Free — 0 USD per user/month, up to 5 seats
  • 5 AI notes and 5 AI recordings for life per user
  • 5 audio/video uploads
  • AI summary
  • transcription
  • action items
  • Ask Fellow
  • and Google Meet
  • Zoom
Business — 15 USD per user/month billed annually (23 USD monthly), up to 50 seats
  • unlimited AI notes and recordings
  • 50 uploads per user per month
  • sales recap templates
  • org-wide templates
  • advanced Salesforce and HubSpot integration
  • keyword tracking
  • bot avatar customisation
Enterprise — 25 USD per user/month billed annually, unlimited seats from a 10-user minimum
  • advanced recording permissions
  • domain control
  • user provisioning
  • security and legal reviews
  • org-wide analytics
  • AI-powered CRM field updates
  • transcript redaction
  • SSO
Special offers — Startups programme: up to 25% off the first year, on application through the vendor's form · Discounted rates for non-profit organisations · Discounted rates for staff and faculty members of educational institutions · 14-day free trial of the Fellow Pro plan with AI recording credits, no credit card required · Annual billing discount: Team drops from 11 to 7 USD per user/month, Business from 23 to 15 USD
Prices and plans listed above may evolve. Always check the official pricing page before subscribing.
Trust & Privacy

Data, GDPR & hosting

A consolidated view of how Fellow handles your data.

GDPR overview

GDPR treatment is concrete rather than declarative. The privacy policy carries a dedicated EEA/UK/Swiss addendum last updated 5 March 2026, naming Fellow Insights, Inc. in Ottawa as controller and citing legal bases article by article: 6(1)(b) for contract performance, 6(1)(f) for legitimate interests, 6(1)(a) for consent, and 9(2)(a) for biometric voice data. Transfers outside the EEA rely on adequacy decisions (Art. 45), Standard Contractual Clauses and the UK IDTA (Art. 46), plus supplementary measures. Access, rectification, erasure, restriction, portability, objection and consent withdrawal are all set out, with a link to the EDPB list of supervisory authorities. Fellow's help centre confirms an on-staff Data Protection Officer and GDPR-compliant data processing agreements. One gap: no Article 27 EU representative is named, and the company has no EU entity or address.

Who owns the data?

When an organisation subscribes, Fellow acts as data processor and the customer organisation is the data controller; Fellow is controller only for its own operations — running and securing the service, usage analytics, billing, legal duties and, with explicit consent, voice samples. Recordings and AI notes are visible by default to internal meeting attendees only, governed by role-based permissions and information-barrier policies between business units. Organisations do not see the content of members' notes or feedback by default, though Fellow will comply if the organisation's own policy demands access. Sharing runs through permissioned channels and is access-logged. In an acquisition, bankruptcy or change of control, personal data may pass to the acquiring entity.

Reuse rights

Customers keep normal use of their own content: notes, transcripts, summaries and action items can be exported, shared through permissioned channels and pushed to connected tools such as Salesforce, HubSpot, Jira, Asana or Notion without asking Fellow for permission. On Fellow's side the stated uses are account creation, service delivery and improvement, support, transactional and security messages, smart suggestions and — with consent — marketing. Anonymised data feeds product analytics and aggregate reports shared back with the organisation. The hard limit is training: customer data is not used to train AI models, data sent to an AI vendor is deleted immediately, and contractual provisions with every AI vendor forbid training on Fellow customer content. Workspace administrators can switch the AI sub-processors off entirely, which disables the features that depend on them. Voice samples for speaker labelling are opt-out at user level and can be disabled workspace-wide, and are never collected from external attendees.

Data retention & training

Retention summary
Fellow keeps personal data as long as it is needed for the purpose it was collected, or to meet legal obligations, settle disputes and enforce agreements. Retention is configurable per workspace, from a set number of days to indefinite, with separate schedules for recordings and transcripts and no user-level override. Optional Zero-Day Retention deletes source recordings and transcripts immediately after processing while keeping AI summaries, decisions and action items. On request, an organisational account is deleted within 15 days, with backup servers purged within a further 30 days. AWS backups are held 30 days and API audit logs 90 days. Voice samples are erased without undue delay if consent is withdrawn. Feedback exchanged between several people cannot be deleted individually.
Trains on customer data
No
Subprocessors disclosed
Yes
DPA available
Yes
GDPR contact

Hosting summary

Fellow hosts its application infrastructure on Amazon Web Services in the Canada Central region, with backups kept for 30 days. Its public sub-processor list, last updated 17 June 2026, confirms Canadian data residency for the three infrastructure providers that hold customer content — AWS, Elasticsearch and PlanetScale — while noting head offices in the United States and the Netherlands. Everything else sits in the United States: Cloudflare for DNS and caching, Datadog, Sentry, SendGrid, Stripe, Intercom, Clearbit, HubSpot, Heap Analytics and Segment. The optional AI sub-processors are also US-based: Anthropic, OpenAI, Google, Groq, Together.ai, AssemblyAI, Deepgram, Recall AI and Fly.io, all disabled if an administrator turns the AI toggle off. Data is encrypted with AES-256 at rest and TLS 1.2+ in transit, and voice samples are stored encrypted in the primary data centre and never shared with third parties. Transfers out of the EEA rely on adequacy decisions and Standard Contractual Clauses. The legal jurisdiction is Ontario, Canada.

Hosting countries
🇨🇦 Canada
Watch-outs

Things to keep in mind

Risks and trade-offs to weigh before adopting Fellow.

  • Recording every conversation builds a searchable archive that can be subpoenaed, breached or read by an admin later — decide your retention policy before you decide your note-taking tool
  • The free plan's five AI notes and five recordings are lifetime, not monthly: teams often discover this only after the quota is gone
  • Consent is only partly automated. The bot announces itself, but with botless desktop capture the duty to tell participants they are recorded falls entirely on the organiser
  • Voice samples for automatic speaker labelling are biometric data; the feature is opt-out rather than opt-in, so users must actively disable voice detection if they object
  • Leaning on AI recaps erodes the habit of listening and writing things down — the summary becomes the meeting, and nuance, disagreement and tone quietly disappear from the record
  • Enterprise-grade safeguards are advertised throughout the marketing but gated behind the Enterprise plan: buyers on Team or Business get the notetaker without the governance they read about
  • Compliance claims rest on documents you cannot inspect up front — the SOC 2 report needs an NDA and the trust centre needs JavaScript, so due diligence has to be requested rather than read
Setup

Setup & Integrations

Technical difficulty

Low for end users. Sign-up is self-service with no credit card, and the only mandatory technical step is connecting a calendar — Google Workspace, Office 365 or iCloud. Google or Office 365 SSO is the recommended login. Botless capture needs a macOS or Windows desktop install; iOS, Android and Chrome clients are optional. No development skills are required for daily use, and the API is optional from the Team plan up. Administrators face more work: recording policies, retention schedules, redaction rules and the AI toggle all live in Workspace Settings. An onboarding programme and a dedicated account manager come with Business and above.

Deployment

Web appDesktop appMobile appIOS appAndroid appChrome extensionBrowser extensionSlack appAPI

Apps stores

Integrations

Asana BambooHR Box Claude ClickUp Confluence Dropbox Figma Freshworks GitHub Glean Global Relay Google Calendar Google Drive Google Meet Google Tasks Google Workspace HubSpot ICloud Jira Linear Loom Microsoft Office 365 Microsoft Teams Microsoft To Do Miro Monday.com Notion Okta OneDrive OneLogin Pipedrive Salesforce Slack Spotify Todoist Trello Vidyard Vimeo Workday YouTube Zapier Zendesk Zoom
Company

Behind Fellow

Company name
Fellow Insights Inc.
Founded
31/10/2018
Country of origin
🇨🇦 Canada
Headquarters
532 Montreal Rd #275, Ottawa, ON K1K 4R4, Canada
UBO
INFORMATION_NOT_FOUND
UBO country
INFORMATION_NOT_FOUND
Domain registrar country
🇺🇸 United States
Support contact

Fundraising

2019 — 6.5 M USD seed round, announced alongside the public launch (TechCrunch coverage of 27 June 2019 linked from the About page)
2021 — 24 M USD Series A, announced with Fellow 3.0 (TechCrunch coverage of 27 October 2021 linked from the About page)
Total announced to date: 30.5 M USD across two rounds; no funding after 2021 is mentioned on the site

Social

Official links

Resources

All the official URLs gathered for verification and reference.

Compare

Alternatives

Tools that compete with or complement Fellow.

A ArvoF FathomF FirefliesG GranolaJ JamieO Otter
FAQ

Frequently asked questions

Does Fellow train its AI models on my meeting data?
No. Fellow states in three separate places that customer data is never used to train AI models. Data sent to an AI vendor is deleted immediately within the LLM, and contractual provisions with every AI vendor forbid training on Fellow customer content. Workspace administrators can also switch off the optional AI sub-processors altogether from Workspace Settings, which disables the features that depend on them.
What certifications does Fellow hold?
Fellow claims SOC 2 Type II, with audits by AICPA-accredited third parties running since 2020, and the latest report covering Security and Confidentiality is available to current and prospective customers on request under a mutual NDA. It also declares HIPAA compliance with a Business Associate Agreement on request, and GDPR compliance. Encryption is AES-256 at rest and TLS 1.2+ in transit, with annual third-party penetration tests.
Where is my data hosted?
Fellow hosts its infrastructure on Amazon Web Services in the Canada Central region, with backups retained for 30 days. Its published sub-processor list, last updated 17 June 2026, confirms that AWS, Elasticsearch and PlanetScale hold data in Canada, while support, analytics and the optional AI sub-processors — Anthropic, OpenAI, Google, Groq, Together.ai, AssemblyAI, Deepgram, Recall AI and Fly.io — are located in the United States.
Can Fellow record a meeting without sending a bot into the call?
Yes. The native macOS and Windows desktop app captures audio and video directly, so no bot joins the meeting, and the iOS and Android apps record in-person conversations from a phone. Bot-based recording remains available, so teams choose per meeting. Both modes run under the same admin recording policies, redaction rules and retention settings.
Who can see my recordings and AI notes?
By default only internal meeting attendees, with no setup required. Access is governed by role-based permissions and information-barrier policies that keep notes siloed between teams and business units. When you do share, it happens through permissioned channels scoped either to the whole workspace or to a specific group, and every share is admin-governed and access-logged.
Is there a genuinely free plan?
There is a permanent free plan at 0 USD, but the allowance is five AI notes and five AI recordings per user for the lifetime of the account rather than per month, capped at five seats. Separately, new workspaces get a 14-day free trial of the Fellow Pro plan with AI recording credits, after which they are downgraded automatically to the free plan.
Does Fellow have an API?
Yes. Fellow publishes a REST API documented at developers.fellow.ai, with workspace owners deciding when it is enabled, revoking user keys at will and seeing a last-used timestamp for each key. Every request is retained in a 90-day audit log. Fellow also ships an MCP server and a Claude connector. API access starts on the paid Team plan, not the free plan.
How long does Fellow keep my data, and can I get it deleted?
Retention is configurable at workspace level, from a set number of days to indefinite, with separate schedules for recordings and transcripts. Optional Zero-Day Retention deletes source recordings and transcripts right after processing while keeping AI summaries. On request, an organisational account is deleted within 15 days, with backups purged within a further 30 days. Deletion requests go to the privacy address in the policy.
How many languages does Fellow transcribe?
The site is inconsistent: the homepage FAQ says 90+ languages, the enterprise FAQ says automatic language detection and transcription for 92 languages, and the pricing feature table says 99 languages. No list of the actual languages is published anywhere, so the exact coverage should be confirmed with the vendor before committing.
Are there discounts beyond the standard plans?
Yes. Fellow runs a startups programme offering up to 25% off the first year on application, and advertises discounted rates for non-profit organisations as well as staff and faculty at educational institutions. Annual billing itself is the largest saving: Team drops from 11 to 7 USD per user per month, and Business from 23 to 15 USD.
Conclusion

Should you pick Fellow?

Fellow is one of the more serious answers to a question the AI notetaker market usually skips: what happens to the recording afterwards. Nine years of product work show. The capture itself is competent rather than unique — bot or botless, across Zoom, Teams, Google Meet, Webex and in-person rooms, with summaries, action items and CRM sync that most competitors also claim. What is genuinely differentiated is the layer above it: recording policies by team or counterparty domain, information barriers, pause events written to the log, redaction that reaches the audio as well as the transcript, Zero-Day Retention, and deletion trails exportable through an admin API. For a compliance officer at a hedge fund or a law firm, that list is the product.

The transparency around it is above average for the category. The sub-processor page is dated, names data locations and separates the optional AI vendors from the rest; the help centre states plainly that infrastructure runs on AWS Canada Central with 30-day backups. The commitment that customer data never trains AI models is repeated consistently rather than buried.

The reservations are commercial rather than technical. The free plan — five AI notes and five recordings for life — is a demonstration, not a tier, and the safeguards used to sell the product hardest, including redaction, advanced permissions, SSO and the HIPAA BAA, all sit behind Enterprise pricing from ten seats up. Smaller buyers get the notetaker, not the governance. Two smaller irritants: the transcription language count contradicts itself across three pages, and the SOC 2 report only appears under NDA, with the trust centre unreadable without JavaScript. Worth a close look if meeting records carry regulatory weight; less compelling if you simply want notes.