All sectors pilot v0.1.0

Meeting Summarizer

Turns a raw meeting transcript (Teams, Zoom, locally-transcribed audio) into a structured CR: decisions made, action items with owner and deadline, open questions.

See the live demo Source code on GitHub

Measured gain

20-40 min

par réunion

Across 50 weekly meetings in a mid-size company, that's 30-60 hours of memo writing avoided. More importantly: no action item gets lost, because the agent doesn't procrastinate.

The problem

Meeting minutes are the most expensive blind spot in the company.

Today

Compte-rendu manuel (ou pas)

Le résultat le plus fréquent : pas de compte-rendu

45 min / réunion COMEX

× 50 réunions / semaine

→ 30 à 60 h de rédaction évitées (ou oubliées)

· Otter / Fireflies produisent une transcription brute, pas un compte-rendu
· Personne n'a le temps de rédiger les décisions actées
· Actions sans porteur explicite ni échéance
· Les COMEX sensibles ne passent pas dans le cloud

0 compte-rendu fiable la plupart du temps

With Meeting Summarizer

Compte-rendu structuré

Transcription locale → CR exploitable, aucune sortie de données

1-2 min / réunion

0 PII vers un cloud externe

→ ~50 h récupérées chaque semaine

› Décisions effectivement actées (vs simplement évoquées)
› Actions avec porteur + échéance ; « À déterminer » plutôt qu'inventer
› Détection RH / M&A / médical → mention « Diffusion restreinte »
› Tourne sur l'appliance LMbox, aucune donnée ne sort

Aucune action n'est plus oubliée

How it works

Four steps, from incoming document to human decision.

1

Transcript arrives

From the local transcription system (Whisper on the LMbox) or Teams / Zoom export. No raw audio needed or stored.
2

Identify participants + structure

Extracts named speakers (or keeps 'Speaker 1', 'Speaker 2' if anonymous), detects date and title, classifies the meeting (decision-making / exploratory / status update).
3

Extract decisions + action items

Actually-made decisions (vs. merely mentioned) listed explicitly. Action items extracted with owner + deadline; if either is missing, the agent writes 'TBD' rather than inventing.
4

4-section memo + sensitivity flag

Summary / decisions / actions / open questions. If content is sensitive (HR, M&A, healthcare), the agent adds a 'Restricted distribution' notice at the top.

Architecture

Tools, connectors, deployment.

Tools (function-calling)

No external tools - direct analysis on the input document.

Optional connectors

outlook
jira
slack

Each connector activates based on the customer's subscription.

Deployment via the SDK

$ lmbox agent deploy ./meeting-summarizer \
    --box BOX-XXX \
    --token "$LMBOX_BOX_API_KEY" \
    --api https://api.lmbox.ai

LMbox guarantees across the catalogue

Data stays with you

Model and data stay on the customer's LMbox appliance. No patient, contract or invoice data is ever sent to an external cloud.

Audit chain

Every tool call, every agent output is timestamped, hashed and admissible before the regulator (ACPR, ANSM, CNIL, EBA).

Human decision

The agent recommends, the human decides. No auto-signature, no auto-payment: final responsibility stays with the business.

Other catalogue agents

Law firms pilot

NDA Reviewer

Reads each incoming NDA, identifies non-standard clauses against the firm's internal template library, drafts an amendment memo for the responsible partner.

1 h 30

Measured gain · par contrat

2 tools 2 connectors

v0.1.0 See the agent

Banking · Insurance pilot

Continuously monitors a panel of validated suppliers / counterparties. Detects UBO changes, OFAC / EU / UN sanctions, litigation. Generates a weekly report for the RSSI plus real-time alerts on critical drift.

Measured gain · faux positifs vs revue manuelle

2 tools 2 connectors

v0.1.0 See the agent

Try Meeting Summarizer on the public demo.

One-click sign-in. You see the agent installed on a real LMbox, with its system prompt loaded and audit chain live.

Open the demo Become a distribution partner