Case study · Built by Strat & Story

Scout


An AI agent that covers the sessions its owner can’t attend live.

Good thinking now comes from everywhere: conferences, webinars, livestreams, podcasts, product launches. The hard part is getting to it. The best events run parallel tracks, the keynote lands at 3 a.m. in another time zone, and the webinar is scheduled against a client meeting. Choosing one session means missing three others.

Scout covers the sessions that can’t be attended live. It joins only events its owner has registered for, captures the talks and the slides, and flags the ones worth going back to watch in full. The workflow has two steps: register for the event, and let Scout take it from there.

Capture
1
Scout
→
Google CalendarIdentifies upcoming events
→
Custom dashboardPresents events for confirmation
2
Scout
→
Auto-joinOpens the link, handles sign-in, enters at start time
Guardrails keep it out of person-to-person meetings.
3
Scout
→
Chrome extensionCustom-built, runs in the browser tab
→
audio
DeepgramReal-time speech-to-text
video
Claude VisionSlide capture and extraction
↓ transcript + slides
Evaluate
Claude triage Judges each session against the owner’s professional context
Gold Watch in full Worth a look
↓ gold items only
Learn
Deep analysis Six-lens extraction
→
Vectorized database Tagged by topic, framework and speaker
→
AI workteam Shared through the Model Context Protocol
→
Auto-surfaces In client briefings

Scout joins only events its owner has registered for. If anything unexpected happens at the door, such as a waiting room, a challenge or an unfamiliar form, it leaves rather than pose as a person. Auto-join is opt-in, there is no hosted service, and captures are stored on the owner’s own machine.

Scout is one member of a multi-agent AI team. Other agents handle storage, synthesis, auditing and surfacing. They work together directly through the Model Context Protocol, with no file drops or manual hand-offs.

At a three-day virtual conference, Scout covered all 20+ sessions across the parallel tracks (about 18 hours of content, more than one person could attend) and delivered:

9
deployable frameworks, with direct bearing on client work
15+
key insights
6
topic areas flagged for deeper study

The result was one brief tuned to the practice, plus a short list of sessions worth watching again in full.

Scout has now captured more than 30 sessions, at events including Dreamforce, Opticon, ZapConnect and IIEX AI.

Scout started as a fix for our own problem. It also shows how we approach AI with clients: start with a real piece of work, build something that does it, and measure what it returns. The same approach runs through our AI strategy work and through Private AI Coaching.

Questions about Scout, or about building something like it:

Schedule a conversation

Or email hello@stratandstory.com

The capture layer is open source: Scout on GitHub →