Steno, Benchmark's self-hosted, open source AI notetaker, capturing meeting notes and action items without any data leaving the firm's own infrastructure.
A leading US-based professional services firm advises clients in banking, insurance, and healthcare.
The firm wanted to equip its consultants with an AI-powered meeting assistant that could securely capture meeting notes, generate summaries, and track follow-up actions while ensuring all conversation data remained entirely within its own infrastructure.
The firm runs distributed teams across multiple offices and time zones, with most meetings held on Google Meet, Zoom, and Teams.
Every project depends on accurate records of decisions, commitments, and action items, yet note quality varied from person to person and important details were regularly lost once a call ended.
There was no searchable record across meetings, so teams spent hours reconstructing what was agreed weeks earlier.
Compliance and client confidentiality requirements ruled out cloud recording services, leaving the firm stuck between manual notes and unacceptable privacy risk.
Every operational bottleneck reported was matched to the workstream(s) that resolved it.
Consultants spent significant time writing and cleaning up meeting notes instead of doing billable work.
Claude turns raw transcripts into clear summaries, decisions, action items with source quotes, topics, and next steps.
Commitments made on calls were often forgotten or disputed later because there was no reliable record.
Claude turns raw transcripts into clear summaries, decisions, action items with source quotes, topics, and next steps.
Popular AI notetaker services store recordings on external servers, which conflicted with client confidentiality obligations.
The entire stack, built on Next.js, PostgreSQL, Redis, and MinIO, runs inside the firm's own environment, so meeting data never leaves the company.
There was no single place to search across past meetings, so knowledge stayed locked inside individual notebooks.
Users can chat with a single meeting or their entire meeting history, with global search and keyword trackers.
Recordings are linked to a click to seek transcript, with export to txt, json, csv, and pdf.
Every team documented meetings differently, making handovers and audits harder than they needed to be.
A bot joins Google Meet calls from the calendar, records them, and transcribes every conversation. Zoom and Teams support is on the roadmap.
Claude turns raw transcripts into clear summaries, decisions, action items with source quotes, topics, and next steps.
Consultants stopped writing notes by hand and reviewed AI generated summaries instead.
Every decision and action item was captured with source quotes, ending disputes about what was agreed.
Meeting data stayed on the firm's own servers, satisfying client confidentiality requirements.
Teams could search across months of meetings in seconds, making handovers and audits faster.
Standard note templates gave every team the same structure and quality of records.
Time to find past decisions likely dropped from hours to minutes through cross meeting search and Ask AI.
