Meeting intelligence with accountable follow-through

The workflow

Meeting transcripts can arrive as text, caption files, office documents, or selected files from Google Drive. The application normalizes the content, records the source context, and submits the transcript to a server-side extraction endpoint.

The extraction step identifies genuine decisions, commitments, risks, issues, and follow-ups. Each proposed item remains visible with its classification and supporting context so a person can review it before relying on it operationally.

Implemented functionality

• Transcript intake for plain text, caption, document, presentation, PDF, and rich-text formats.
• Google Drive file selection using read-only access when the integration is configured.
• Structured extraction of action title, description, stakeholder role, tracking level, priority, due date, executive office, and insight category.
• A per-item reasoning field that provides review context for the proposed classification.
• Optional creation of a Salesforce task after authentication, carrying the action description, status, priority, due date, and meeting-intelligence context.

Human control

The system proposes structured actions; it does not make final business decisions. Authentication is required for connected systems, Google Drive access is read-only, and task handoff is a distinct step rather than an invisible side effect of transcript processing.

Scope and limitations

This is an application-engineering and AI-assisted workflow example. It does not claim perfect transcription, perfect extraction, automatic assignment acceptance, measured productivity gains, organization-wide adoption, or unattended decision-making. Long transcripts may be clipped before extraction, and connected services require their own configuration and authorization.

Privacy

Participant names, organizations, email addresses, meeting links, dates, locations, transcript excerpts, action contents, record identifiers, and credentials are omitted. The example explains the implemented pattern without reproducing private meetings or customer records.