Day 3 — Enterprise patterns, observability, security, takeaways
Goals for the day
- Move from “working agent” to production-grade agent
- Surface the cross-cutting concerns NBS will hit: security, observability, cost, deployment
- Capture takeaways and next steps for the NBS engineering org
Agenda
| Time | Session | Format |
|---|---|---|
| 09:30 – 09:45 | Day 2 recap | Discussion |
| 09:45 – 11:00 | Production architecture — deployment topologies, scaling, async patterns | Lecture + reference architecture |
| 11:00 – 11:15 | Break | — |
| 11:15 – 12:30 | Observability — tracing agent runs, debugging loops, replay | Lecture + demo |
| 12:30 – 13:30 | Lunch | — |
| 13:30 – 14:30 | Security — prompt injection, tool authorization, secret handling | Lecture + threat model walkthrough |
| 14:30 – 15:30 | Cost & latency — token economics, caching, batching, model routing | Lecture + numbers from real deployments |
| 15:30 – 15:45 | Break | — |
| 15:45 – 16:30 | Lab 3 — take one of your Day 2 agents, harden it for prod | Hands-on |
| 16:30 – 17:00 | Workshop wrap — takeaways, open questions, recommended next steps for NBS | Group discussion |
What participants will leave with
- A reference architecture they can adapt for NBS internal services
- A hardened version of one agent (auth, logging, eval-gated deploy)
- A shortlist of internal NBS use cases worth piloting
Materials
- Reference architecture diagrams: link TBD
- Observability stack walkthrough: link TBD
- Security checklist (will be expanded on Day 5 in the guardrails session)
Attendees
- NBS: TBD
- MetaWeave: TBD
Decisions
- Captured live.
Action items
- Owner — action — due
Bridge to Day 4
The workshop ends here. Day 4 shifts mode entirely — from teaching to listening. MetaWeave reviews NBS applications and shares an opinionated read.