Fathom AI Notetaker vs Leexi

Fathom AI Notetaker

The captured pages show a bot joining Zoom, Google Meet, Microsoft Teams and Slack Huddles, plus a documented bot-free Mac beta that captures device audio while keeping speaker attribution and mute acknowledgment, with capture, transcription and summarization in real time. But we found no public information on the supported languages — German is never confirmed — nor on custom vocabulary, timestamps, transcript editing or file upload, and the dedicated iOS app for in-person meetings is announced as not live yet, so my workshop-in-a-client-room use case stays unanswered.

Leexi

The assistant joins Teams, Meet and Zoom among ten conferencing tools including Webex, records in-person workshops from an iOS or Android app with no room hardware, and pulls calls from eleven VoIP services — every channel my week produces. Diarization, a company dictionary for jargon, acronyms and project names, accent adaptation, and live notes through a web extension and dashboard captioning are documented, and the speech-to-text providers are named. What I miss as a German-reading buyer: the pages claim over a hundred languages without naming German or publishing accuracy per language.

Fathom AI Notetaker

Decisions and action items assigned by name, a three-sentence summary, follow-up emails, search and questions across team calls, and sharing to Slack, Asana or Notion are all documented. We found no public information on templates per meeting type, and the API documentation states summaries are "Always displayed in English" — for a practice that runs half its meetings in German, minutes that are English-only as documented are close to disqualifying.

Leexi

Minutes adapt by meeting type — steering committee, workshop, interview — with customizable templates, and the AI extracts decisions, tasks with suggested owners and deadlines, which is exactly the protocol I write by hand each night. Ask Leexi searches the archive, minutes go out by email and export to Word, Google Docs or PDF, and the language models behind them are named. We found no public information on answers that cite the underlying transcript passage, nor on selecting the summary language.

Fathom AI Notetaker

Google and Outlook calendar connection (required, read-only), native HubSpot and Salesforce sync with AI-extracted fields mapped to records, Slack, Asana, Notion, Zapier and Make, a documented REST API with webhooks and TypeScript/Python SDKs, and an official MCP server for ChatGPT and Claude — the strongest part of the file. We found no public information on Microsoft Dynamics or Pipedrive, Confluence or SharePoint export, or recording rules per calendar or meeting type.

Leexi

Google, Outlook and on-premise Exchange calendars drive automatic joins, eleven CRMs including Salesforce, HubSpot, Pipedrive and Dynamics receive notes automatically, and Slack, Notion, Google Docs, SharePoint, Zapier and Make are documented alongside webhooks. A public REST API with reference documentation and a read-only, tenant-scoped MCP server with OAuth close the loop for our own tooling. We found no public information on API rate limits or versioning, and CRM field mapping per meeting type is not described.

Fathom AI Notetaker

The terms require that no participant be recorded without consent, the bot-free mode preserves mute acknowledgment, and the data processing agreement promises assistance with data protection impact assessments. Beyond that we found no public information on participant notices, an opt-out mechanism, consent logging, or the defaults for the coaching metrics and AI scorecards — and no works-council pack or works-agreement template, which is what I would need before introducing this to our team.

Leexi

Dedicated pages walk through recording law — participants must be informed before recording begins, CNIL bars permanent systematic recording, the AI Act transparency article is cited, and admins control when the assistant joins automatically. That is a real basis for a Betriebsrat conversation. But we found no public information on a participant opt-out mechanism, a works-agreement template or DPIA support, and the coaching module's evaluation scorecards and performance analysis are precisely the per-person analytics a works council will interrogate, with no documented way to switch them off.

Fathom AI Notetaker

SOC 2 Type II, encryption at rest and in transit, SSO with SCIM, admin-level retention controls with regional storage options and audit logs, deletion on request, and custom retention are all stated. The training promise is qualified rather than absolute: third parties such as OpenAI or Anthropic are barred from training, but Fathom trains its own in-house models on de-identified content unless you opt out — not the written guarantee my clients ask for — and we found no public information on discarding audio after transcription, legal hold, or export of the full archive.

Leexi

The promise I want in writing is there: no training on conversations on any account, stated in the privacy policy, the compliance FAQ and a dedicated no-training attestation, though the commitment sits outside the DPA text itself. Retention defaults to a maximum of two years, reducible by the client in platform settings, with per-plan audio limits, single sign-on with SCIM, eight roles, audit logs, encryption in transit and at rest, and ISO 27001:2022, plus ten named subprocessors with their regions. Whole-archive export is manual rather than one click, and the vendor itself states SecNumCloud and health-data certification are not yet held.

Fathom AI Notetaker

The contracting entity is Fathom Video Inc., a Delaware corporation in San Francisco, the service is stated to be hosted in the United States, and the privacy policy names EU and UK representatives in Cork and London with Standard Contractual Clauses and the Data Privacy Framework as safeguards; a data processing agreement is published. The data processing agreement refers to a subprocessor list at the trust center, but we found no public information naming those providers or their locations, so nothing in the chain is evidenced as European.

Leexi

A Belgian company with its seat in Brussels hosts recordings, transcripts and minutes on AWS in Paris, with a Scaleway option above certain licence volumes for buyers wanting an operator beyond extraterritorial reach. The language model is OpenAI run via Azure France in a closed circuit, Mistral serves as a European model, speech-to-text goes to Gladia, ElevenLabs and Azure Speech, and ten subprocessors are published with locations next to a public DPA. The compliance FAQ itself notes AWS is a US company theoretically subject to the Cloud Act, so the chain is European-hosted end to end but not European-operated in every layer.

Fathom AI Notetaker

Prices are public and simple: a free plan with unlimited recordings and no usage caps, Team at $15 /month / per user (2 user min), Business at $25 /month / per user (2 user min), Enterprise at $35 /month / per user annual only, with annual billing saving 25%+, prices in US dollars and taxes borne by the customer. The Launch Assist implementation service is marked only as a paid add-on without a figure, and we found no public information on what exactly counts as a seat, storage limits, or VAT treatment — so I cannot compute a full annual cost for my twelve-person firm from the pages alone.

Leexi

Three tiers carry published per-licence monthly prices — twelve, twenty-three and thirty-nine euros, discounted twenty percent from fifteen, twenty-nine and forty-nine — on annual invoice excluding VAT, with minimum users stated, volume discounts from ten licences and a seven-day free trial. But single sign-on, SCIM and the governance module sit in the Enterprise tier priced only on contact, and the sovereign Scaleway switch and the consultant service carry no public price. What counts as a seat beyond 'per licence' and overage handling are not described.