Meeting intelligence · for product teams

Frame the meeting.
Surface the decision.

Meridian Note listens to your product reviews, sprint planning, customer interviews, and exec syncs — then returns a structured summary, decisions log, and action items routed to Linear, Notion, or Slack. No transcripts. Signal only.

Request access See how it works ↓

Currently onboarding teams of 8–60 PMs · EU data residency · Beta access only

What lands in your inbox

Three artefacts. No verbatim transcripts.

A.

Decision log

What was decided, what was deferred, what was rejected — and the dissent that almost moved it. Linked to the moment in the call where it happened.

B.

Structured summary

Topics covered, time spent on each, who drove which thread. Optimised for skimming on the way to the next meeting.

C.

Action items, routed

Owner-attributed, deadline-extracted, pushed straight into Linear / Notion / Slack with the originating context attached. No copy-paste.

Why product teams pick Meridian Note

Built for the way PMs already run meetings.

01

Speaker-aware, not speaker-blind

Pull participant names from your calendar invite — get attributed action items instead of "Speaker 3 said…" Works on Zoom, Google Meet, Teams, and uploaded recordings.

02

Decision-graded, not topic-graded

Most tools surface what was *discussed*. Meridian surfaces what was *decided* — and tags the moments where commitment hardened, softened, or got punted to the next sprint.

03

Customer-interview mode

Toggle interview mode and Meridian extracts pain quotes verbatim, codes them against your hypothesis ladder, and flags moments where the user said one thing and meant another.

04

Routes into your stack — properly

Linear (issue + cycle + parent), Notion (page in your meeting database with the right relations), Slack (DM the owner, post summary in #product-team). Two-way sync on status, not fire-and-forget.

05

EU-only processing, by default

Audio + transcripts + LLM inference all run in EU (Paris + Frankfurt) regions. CNIL-compliant retention. No US sub-processors for any audio/text payload. Data residency in writing on the contract.

From the calendar to the action item

Four steps, one minute of attention from you.

STEP 01

Connect calendar + recording source

Google or Microsoft calendar. Zoom Cloud Recording, Google Meet, or Teams. We auto-discover meetings tagged with our keyword, or pick the meetings you want covered.
STEP 02

Audio routes to EU processing

Encrypted in transit + at rest. Whisper (large-v3) for transcription, Mistral / Llama 3.1 70B for structuring — running in OVH Roubaix or Scaleway Paris. No leaving the EU.
STEP 03

Decision log + summary in your inbox · ~6 min after the call ends

Email + Slack DM. Action items already created in Linear or Notion if you've connected them. Nothing to file manually.
STEP 04

Audio deleted after 30 days

Transcripts kept until you delete them; raw audio is purged automatically per our default retention policy. Configurable per workspace.

Privacy posture

Built for product teams in regulated environments.

Sub-processors: Whisper (self-hosted on OVHcloud Roubaix) · Mistral (Mistral AI · Paris) · Postgres + S3 (Scaleway Paris). No US-based processor touches your audio or transcript.

Data residency: Storage and inference both in France. The DPA references CNIL guidance on AI processing of professional speech (2024 update). EU SCCs are not invoked because no transfer occurs.

Right of erasure: Workspace admins can purge any meeting, transcript, or summary. We keep audit logs of deletions for 30 days, then those are purged too.

Customer-named entities: Names of individual customers / prospects mentioned in calls are flagged as PII and can be redacted before LLM processing on a per-workspace toggle.

Full privacy posture →

Frequently asked questions

Things product leaders ask before we kick off.

Q.01

Does it work without Cloud Recording? Some of our exec syncs don't auto-record.

+

Yes. We can join the call as a silent bot (Zoom / Meet / Teams) when Cloud Recording isn't enabled. The bot records to our EU-resident audio bucket, never to the meeting host's account. Disable it per-meeting if a participant objects — Meridian asks at the top of the call by default (configurable).

Q.02

How accurate is action-item extraction? Our last tool routed half of "we should explore" as commitments.

+

Our action-item classifier was trained on ~12k labelled product-meeting moments distinguishing commitment from speculation. Internal F1 ≈ 0.84 on commitment detection, with explicit handling of softeners ("we should probably look into…" → flagged as exploration, not action). You can review and reject any item before it's pushed to Linear / Notion — the first week of usage is mostly tuning sensitivity for your team's phrasing.

Q.03

We talk about specific customers by name. Are those names sent to an LLM?

+

Per-workspace toggle. Default is on for B2B SaaS workspaces (since named accounts are usually the point of the conversation), off for consumer / regulated workspaces. When off, named entities matching your customer list are replaced with stable hashed tokens before LLM processing — the resulting summary references "Customer A" / "Customer B" with the mapping kept in your workspace and never sent to inference.

Q.04

Linear integration — does it create issues in the right project, or do we have to triage them later?

+

Per-meeting routing rules. You map each calendar event series → Linear team / project / cycle once, and Meridian respects it. For ad-hoc meetings without a series, we ask the meeting owner where to file it via a Slack button before pushing — usually a 5-second decision rather than a 5-minute triage afterwards.

Q.05

RGPD / CNIL — we're a French company and our DPO will ask. What's the answer?

+

All processing in France (OVHcloud Roubaix and Scaleway Paris-2). DPA signed under French law. CNIL guidance on AI-mediated speech processing (Mar 2024 update) is the framework we follow. Sub-processor list is short and fully European. We've onboarded three regulated FR employers so far — happy to share the DPO-facing memo on first call.

Q.06

Pricing? You don't list it.

+

€18/PM/month annual or €22 monthly. Includes unlimited meetings, all integrations, EU residency, and the per-workspace PII redaction layer. Volume discount above 25 seats. We're explicitly priced below US incumbents because the EU-only processing is the differentiator, not a luxury tier.

Request access

Two questions to know if we're a fit.

We're onboarding teams of 8–60 PMs through Q3 2026. Tell us a bit about yours and we'll come back within one business day with either an invite or an honest "not yet — here's why".

· EU-resident processing
· Linear, Notion, Slack integrations live
· English + French speech today; ES + DE in beta
· Workspace SSO via Google / Microsoft

Frame the meeting. Surface the decision.