How a Business Owner Running Multiple Revenue Streams Got a Second Brain That Captures Voice, Calls, and Meetings, Routes Every Item Automatically, and Never Sends Sensitive Audio Off His Own Machine
Information arrived everywhere: voice memos, phone calls, meeting transcripts, screenshots. It was captured and trusted nowhere. And the owner had a hard line: sensitive business conversations could not go to a cloud transcription service. Epiphany Dynamics built a private second brain with 13 workflows, local transcription on the owner's own workstation, automatic classification and routing into the task and file systems he already used, and three scheduled digests. At delivery, every workflow was live with zero error executions.
Cloud transcription is cheap, $0.21 to $0.36 per audio hour, and it was off the table. The owner's requirement, stated on the first call, was that sensitive captures transcribe and classify entirely on his own workstation. That is not an unusual instinct: in Cisco's 2025 benchmark of 2,600 privacy and security professionals, 64% worried generative AI could leak sensitive information and 90% saw local storage as inherently safer. The local model was not a feature. It was the reason the system could exist.
-
01 Voice memos
Recorded on the owner's phone, transcribed locally, classified, routed.
-
02 Phone calls
Captured, transcribed locally, summarized with action items.
-
03 Meetings
Transcripts ingested, decisions and tasks extracted.
-
04 Screenshots and files
Dropped into a watched inbox, classified, filed.
-
05 Routing
Every capture lands in the task manager as a task or in the file store as a document, with the owner's project structure preserved.
-
06 Digests
A daily summary, an end-of-day close, and a weekly review, each generated on schedule from that day's captures.
McKinsey's Global Institute found interaction workers spend about a fifth of the workweek looking for internal information or tracking down the person who has it. Memory does not close the gap: in a modern replication of Ebbinghaus's forgetting curve, retention was 20% after one day and 10% after six. An owner juggling several businesses, capturing on the go, loses most of what was said by the next morning unless something catches it.
Why local transcription was the scope, not an upgrade
Midway through, the local-model lane could have been framed as an optional add-on. It was not. The owner had asked for it on the first call, and a cloud-only pipeline would have sent every sensitive capture, including the sensitivity check itself, to a third party, which defeats the reason he hired us. The local lane shipped as in-scope at no additional charge. Privacy was the product.
A private client
- Built
- Multi-lane capture, local transcription, classification and routing, three scheduled digests
- Sector
- Data and Knowledge Ops
Published under a pseudonym with the client's permission. A boutique research identity is not used here; this is a private owner running multiple revenue streams.
- Why local transcription?
- Sensitive business conversations could not go to a cloud transcription service. Local transcription on the owner's own workstation was the reason the system could exist.
- What gets captured?
- Voice memos, phone calls, meetings, screenshots and files, each classified and routed into the task and file systems the owner already uses.
- How is it reviewed?
- Three scheduled digests: daily, end of day, and weekly.
Need a second brain that never leaves your machine?
Get a free audit of what should already be captured, routed, and private.
Free 30-minute audit. No obligation, no pitch. Or take the 2-minute AI quiz.