Home/Case Studies/AI Meeting Intelligence
Executive Productivity

Meetings produce decisions. Most tools produce transcripts.

A platform built on the premise that a summary nobody acts on is as useless as no summary at all, so the output is owned actions in the tracker, not a document in a folder.

AI meeting intelligence platform showing decisions and action items
SectorProductivity & collaborationCaptureReal time, during the meetingOutputDecisions, actions, ownersDestinationTask and project toolsGuardrailNothing asserted without a citationPrivacyConsent and retention controls
The situation

The transcript is not the deliverable.

Transcription is solved and largely commoditised. What has not been solved is the reason anybody wanted the transcript: knowing what was decided, what has to happen next, and who agreed to do it.

A raw transcript makes that worse, not better. Reading forty minutes of conversation to extract three actions is more work than the notes it replaced, which is why most meeting tools end up unused after the first month.

The second failure is destination. Actions extracted into the meeting tool are actions in a place nobody looks. Work happens in a tracker, and anything that does not arrive there does not happen.

What we built

Extract decisions, attribute them, and put them where work lives.

Real-time capture during the meeting, not processing afterwards, so the summary exists as the meeting ends and can be corrected while everyone still remembers what they meant.

The extraction is structured rather than narrative: decisions, actions, owners, deadlines and open questions as separate typed outputs. Each one carries a citation back to the moment in the conversation it came from, so a disputed action can be checked in one click instead of argued about.

Where the conversation is ambiguous (and it frequently is), the platform surfaces the ambiguity as an open question, not resolving it by guessing. That is the difference between a tool people correct and a tool people abandon.

Actions are pushed into the systems the organisation already uses. The platform's own record is the audit trail, not the workflow.

Inside the system

What the platform does.

01

Real-time capture

Processing during the meeting, so the summary lands as it ends instead of an hour later.

02

Decision extraction

What was agreed, separated from what was discussed.

03

Action items with owners

Typed actions with an owner and a date, not bullet points.

04

Citations

Every extracted item linked to the moment it came from.

05

Open questions

Ambiguity surfaced, not resolved by guessing.

06

Push to trackers

Actions delivered into the task and project tools people already work in.

07

Consent and retention

Recording consent, retention windows and access controls, because meeting content is sensitive by default.

Built with

Built with.

Speech

Streaming transcriptionSpeaker diarisationReal-time pipeline

Model layer

LLM extractionStructured output schemasCitation binding

Application

Live meeting viewPost-meeting review and correctionSearch across meetings

Integration

CalendarTask and project trackersIdentity and access
What changed

What changed.

  • Meetings end with actions, not homework. The summary exists as the meeting closes and can be corrected in the room.
  • Actions land where work happens. Pushed into the existing tracker rather than held in yet another tool people have to remember to open.
  • Disputes are resolved by citation. Every extracted item links back to the moment in the conversation, which removes the argument about whether something was agreed.
  • Ambiguity is visible. Open questions are surfaced, not silently resolved, which is what keeps the output trustworthy over months instead of days.