Fathom AI Notetaker vs Leexi

Fathom AI Notetaker

A bot that joins Zoom, Google Meet, Microsoft Teams and Slack Huddles is documented next to a bot-free beta (Mac only) that keeps speaker attribution and mute acknowledgment, and capture, transcription and summarisation are described as happening in real time. The captured pages give no supported languages — German is not confirmed — and no Webex, phone or dialer capture, custom vocabulary, timestamps, transcript editing or accuracy figures.

Leexi

The notetaker joins Google Meet, Microsoft Teams, Zoom and Webex, records calls from eleven telephony services, and captures in-person meetings from an iOS or Android app with no room hardware; live captioning appears in the dashboard, a web extension takes live notes, and file upload is listed in the Starter plan. A company dictionary and accent adaptation cover jargon and proper names, and the speech-to-text providers are named, though we found no public information on accuracy per language, mixed-language meetings, or editing of the transcript.

Fathom AI Notetaker

Summaries with decisions and action items assigned by name, search across meetings, Ask Fathom over team calls, trackers, scorecards, a shared workspace and pushes to Slack, Asana and Notion are documented. The API reference, however, states the summary "is always in English", and we found no public information on summary templates per meeting type, answers citing the transcript passage they rest on, or the model that writes the summaries.

Leexi

Minutes adapt to meeting type with organization-defined templates, decisions and actions are extracted with suggested owners and synced into CRM and task tools, and Ask Leexi searches the whole archive, with the MCP server exposing the same calls under the organization's own permission boundaries and the language models named. We found no public information on choosing the summary language, per-language summary quality, or documented hallucination handling and human review.

Fathom AI Notetaker

A documented REST API with webhooks, TypeScript and Python SDKs, OAuth for installable apps and an official MCP server sits alongside native HubSpot and Salesforce field sync, Slack, Asana, Zapier and Make, with the Outlook or Google calendar connection required for the service to work. We found no public information on Microsoft Dynamics or Pipedrive, SharePoint or Confluence export, or recording rules per calendar or meeting type.

Leexi

Calendars including Exchange on-premise trigger auto-join, eleven CRMs including Salesforce, HubSpot, Pipedrive and Microsoft Dynamics receive notes with automated field population, and Slack, Notion, SharePoint, Zapier and Make sit alongside a documented public REST API with reference pages, webhooks and an OAuth-secured MCP server with defined scopes. We found no public information on API rate limits and versioning, or on centrally administered integrations with logs.

Fathom AI Notetaker

The terms prohibit use when any participant withholds consent to be recorded, the DPA promises assistance with data protection impact assessments, and the bot-free capture method is described — points I would credit in a DPIA. But we found no public information on a notice to external guests, a participant opt-out or stop mechanism, works-council material or an EU AI Act classification, and the coaching metrics and AI scorecards carry no documented default state or off-switch.

Leexi

The GDPR pages state that participants must be informed before recording begins, cite the EU AI Act transparency obligation, and give users rights to object and to erasure — but we found no public information on a stop-recording or opt-out mechanism for participants, on defaults and switch-off for the Leexi Coach evaluation scorecards and performance analysis, or on any works-council information pack or works-agreement template. Before a first recording under co-determination I would need every one of those answered in writing.

Fathom AI Notetaker

Enterprise plans carry custom data retention, regional storage options and audit logs, SSO with SCIM, admin sharing rules, encryption at rest and in transit and SOC 2 Type II, and third parties such as OpenAI, Anthropic or Google are barred from training on meeting content — though in-house models train on de-identified data with an account opt-out, and base retention is only "as long as necessary". Voiceprints are collected solely for iOS users who opt in and are classed as biometric under Illinois and Colorado law; we found no public information on the subprocessor names, so the transcription and model providers and their retention stay unverified.

Leexi

The ten subprocessors are named with their region — the reading language model is OpenAI executed via Azure France in a closed circuit plus Mistral in the EU, and speech-to-text runs through Gladia, Eleven Labs and Azure Speech — with a two-year default retention reducible in platform settings, encryption in transit and at rest, ISO 27001:2022 in the DPA, SSO with SCIM and eight roles at Enterprise, audit logs, and return or deletion at contract end. Two things hold me back: the captured pages place the no-training commitment in the privacy policy and compliance FAQ rather than in the data processing agreement itself, and we found no public information on voice profiles, zero retention at the AI providers, discarding audio after transcription, or SOC 2.

Fathom AI Notetaker

Fathom Video Inc., a Delaware corporation under California law, hosts the service in the United States; a data processing agreement on Common Paper standard terms is published, an EU representative sits in Cork, the Netherlands is named for EEA transfers, and SCCs plus the Data Privacy Framework are the stated safeguards. The subprocessor list is only referenced at the trust page, and we found no public information on the names and regions of the subprocessors — including which language model and transcription engine read the recordings — so I cannot verify the chain for a European deployment.

Leexi

Leexi SA is a Belgian company contracting from Brussels, recordings, transcripts and minutes are stored on AWS in Paris with a Scaleway option for stricter sovereignty, every subprocessor is listed with an EU or Europe location, and the data processing agreement is a public document available before account creation. The OpenAI models run through Azure France but OpenAI is not a European operator, the vendor itself notes AWS is theoretically subject to the Cloud Act and that SecNumCloud qualification is not yet held, so a fully European end-to-end chain is not evidenced.

Fathom AI Notetaker

Team at "$15 /month / per user" with a 2-user minimum, Business at "$25 /month / per user", Enterprise at "$35 /month / per user" billed annually only, annual billing "save 25%+", and a free plan with "unlimited recordings with no usage caps" are public, with cancellation without penalty and taxes borne by the customer stated in the terms. We found no public information on what counts as a seat, storage limits, or the price of the Launch Assist add-on, and features such as SSO and custom retention are tier-gated rather than priced.

Leexi

Three tiers are published at "12€/month", "23€/month" and "39€/month" per licence on annual invoice excluding VAT, with minimum users, volume discounts from ten licences, no setup cost and a seven-day trial stated. Single sign-on, SCIM and custom retention sit in an Enterprise tier priced only on contact, and we found no public information on what exactly counts as a seat or on minute and storage limits in the lower tiers.