Read AI

What it is

Read AI indexes and scores conversations across three surfaces — video meetings, email, and chat — instead of only the calls a bot attended. David Shim, Robert Williams, and Elliott Waldron founded it in Seattle in 2021; it has raised $81M, most of it a $50M Series B led by Smash Capital on October 28, 2024 at a reported $450M valuation, and ran roughly 113 staff mid-2026. Capture is bot-based: a Read agent joins Zoom, Google Meet, and Microsoft Teams calls off your calendar and produces a Meeting Report — transcript, summary, action items, playback, and per-participant engagement and sentiment scoring.

The piece that decides stack placement is Ask Read, the search layer over that archive. It answers questions across meeting reports plus connected Gmail and Outlook, Slack and Microsoft Teams chat, Google Drive and OneDrive, Confluence and Notion, and HubSpot and Salesforce. Read ships it to every user, free tier included. A remote MCP server followed on February 25, 2026, and Read AI now sits in Claude’s connector directory, so a Claude chat or agent can browse recent meetings, pull transcripts and meeting details, dispatch a Read agent to record a Teams, Zoom, or Meet call, and share reports out.

Why it shows up in ops stacks

• It indexes email and chat, not just meetings. Otter.ai, Fireflies.ai, Granola, and Fathom archive calls. None of them answers “what did we tell this account over email three weeks ago.” Read spans both. If the retrieval problem crosses channels, that is the one reason to shortlist it.
• Engagement scoring is per participant, not per deal. Read attaches a participation and sentiment score to each attendee of each call. Gong scores deals and forecast risk; that is a different measurement and a different purchase. Per-attendee data is what feeds interview-panel calibration and CSM coaching.
• SSO carries a published price. SAML, SCIM, HIPAA, and custom data retention sit on Enterprise+ at $29.75 a seat annually. Otter puts the same four behind a demo-gated Enterprise tier with no public number, so Read is the one you can budget before the first sales call.
• The MCP server is open beta, not an enterprise SKU. Read’s API and MCP server are available to any user who meets the prerequisites rather than gated behind a custom contract, which is what makes a small pilot cheap.

Pricing

Per user per month from Read AI’s pricing page, checked September 6, 2026. Both annual and monthly rates are published:

• Free — $0. 5 meetings a month, transcripts, summaries, search, meeting coach, 20+ languages.
• Pro — $15 annual, $19.75 monthly. Unlimited meetings, priority processing, 100 file-upload credits a month, premium integrations.
• Enterprise — $22.50 annual, $29.75 monthly. Adds audio and video playback, video highlights, 200 upload credits, premium support. Workspace access requires 5 licenses minimum.
• Enterprise+ — $29.75 annual, $39.75 monthly. Adds HIPAA handling, SAML and SCIM, domain capture, custom data retention, 300 upload credits, workspace onboarding.

Volume pricing runs 10% off at 100 licenses, 15% at 500, and 20% at 1,000.

The number teams land on is $15 annual on Pro: unlimited capture plus cross-channel search, no SSO. That undercuts Otter Business at $19.99 annual and Fireflies Business at $19 annual, and sits $1 over Granola Business at $14. The moment SSO becomes a security requirement the real number is $29.75 — double Pro, and above every annual Business tier in the category. Price the org at the tier your security review will force, not at the tier on the landing page.

Best for

RevOps, Customer Success, and recruiting leads at 25-500 seats who need one searchable record across calls, inboxes, and Slack, and can commit annually at $15 a seat. It is the strongest pick when the question you keep failing to answer crosses channels, and when per-attendee engagement data feeds an actual coaching or panel-calibration program rather than a dashboard nobody opens.

Not for: buyers who want deal scoring and forecast risk — that is Gong, and Read does not replace it. Not for teams under 5 seats that need workspace admin, since the workspace minimum is 5 licenses. Not for orgs whose meetings happen in rooms or whose counterparties decline recording bots.

Alternatives — and when to pick them instead

• Otter.ai — pick it when the project is consolidation rather than a green-field choice. Otter reports 35 million users and 86% of the Fortune 500, so it is normally already expensed somewhere in the building. You give up email and chat indexing, and you inherit an active recording-privacy suit.
• Fathom — the fastest-growing entrant, and the pick when a free tier has to carry real work. Fathom’s free plan records and summarizes without a meeting cap; Read’s stops at 5 meetings a month. A free-tier bake-off goes to Fathom on volume before either product’s quality is tested.
• Fireflies.ai — the closest substitute on price at $19 annual Business. Pick it when the decision turns on meeting-connector breadth alone and cross-channel search is not part of the requirement.
• Granola — pick it when bots get declined or the conversation is in a room. Capture is on-device with no entry in the participant list, and Business is $14. You lose the email and chat index completely, which is the whole reason to buy Read.
• Gong — a different purchase, not a pricier Read. Buy it for deal scoring, forecast risk, and a rep-coaching program with pipeline attached.

Watch-outs

• The free tier cannot carry a pilot. 5 meetings a month is one week for a single CSM or recruiter, and it is not enough to judge transcript quality or engagement scoring against a real calendar. Guard: buy one month of Pro monthly at $19.75 for the pilot cohort and evaluate there. Judging Read on its free tier against Fathom’s uncapped free tier measures the price sheet, not the product.
• Cross-channel indexing is the reason to buy and the widest consent surface you will ever grant a notetaker. Connecting Gmail, Outlook, and Slack hands a vendor the written record too, not just the spoken one. Guard: scope each OAuth grant separately at rollout instead of accepting the full connector set, and get the training-on-your-data clause in the signed DPA rather than the click-through terms. Route the email and chat connectors past whoever owns your retention policy before, not after, the first sync.
• Calendar auto-join sends bots to meetings you did not think about. A blanket auto-join grant puts a recorder in external calls, candidate interviews, and privileged conversations, and the recap lands in front of the full invite list — side comments included. Guard: switch auto-join off and require per-meeting opt-in for the two-party-consent states (California, Pennsylvania, Illinois, Washington, Florida), and set a rule that interview and counsel calls are never on the auto-join list.
• The API and MCP server are open beta, and the exposed tool surface is narrower than the marketing. Connected agents browse meetings, read transcripts and meeting details, start a recording, and share reports. Guard: confirm the exact tool list against your use case before making Read an MCP dependency in a stack — if an agent needs structured action items or engagement metrics rather than transcript text, verify that path works on a trial seat before signing an annual contract.