Building a Team Brain: Connecting Code, Customer Context, and Production Agents
A hands-on technical workshop on building a live product and engineering context layer over your production apps like Zendesk, Datadog, Granola, and Asana.
About the workshop
Most engineering time goes into what to build. The harder, less-discussed question is why build this feature? That customer context is scattered in Zendesk tickets, buried in Datadog traces, sitting in Granola meeting transcripts, siloed in Asana roadmaps. Nobody has time to synthesize it by hand, so most teams don't, and decisions get made on partial context.
Prompting a model is the easy part of building this. The real problems are systems problems: how do you pull live, multi-app customer data securely, manage token overhead across four+ APIs, scope tool execution per user, and turn raw fragments into a context layer you can actually query and trust?
This is the same problem, at different stakes, as shipping a production agent that acts on behalf of your customers. Building an internal context layer forces you to solve per-user scoped access, credential isolation, and token lifecycle management across real third-party systems. Get this right internally, at the scale of your own team's tools, and you've built the muscle for the exact infrastructure a multi-tenant, customer-facing agent needs at 10x the stakes.
We'll move through a Plan → Build → Review workflow, live: wiring real connectors, handling auth and scoping as we go, and shipping a working context layer by the end of the session. Code, not slides.
What you'll build
Connectors into Zendesk, Datadog, Granola, and Asana, each scoped to what the session actually needs. Not blanket access.
A unified context store that normalizes fragmented signals (tickets, metrics, transcripts, roadmap items) into something queryable.
Per-user scoped tool execution, so the context layer respects who's asking, not just what's being asked.
A working retrieval layer your own agents can query. The same interface pattern production agents use to pull customer context on demand.
Who should attend
AI software engineers building production applications beyond a single-user chat box.
Technical PMs and engineering leads looking to bridge the gap between codebase reality and user feedback.
Founders and architects building multi-tenant, agentic workflows in production.
Agenda
4:30 PM - Check-in & networking
5:00 to 7:00 PM - Technical workshop and live build
7:00 – 8:00 PM - Pizza, Q&A, and peer discussion.