AgentOps Value Delivery Workshop
Facilitator-ready, hands-on. This VBD track is a full-day, hands-on workshop. Across seven connected labs, attendees build and operate one real agent on Microsoft Foundry, and every lab consumes the artifact the previous lab produced.
This track expands the one-hour AgentOps Briefing into an eight-hour Value Delivery Workshop. Attendees take one agent - the Contoso Travel Agent - all the way around the four AgentOps pillars on Microsoft Foundry:
Evaluate -> Ship -> Observe -> Operate
The workshop is delivered as a sequence of hands-on labs. Attendees leave with a working end-to-end AgentOps pipeline and an evidence pack, not just a conceptual overview.
Full-day outcome
By the end of the VBD workshop, attendees have built and operated one AI agent in production:
- Deploy the agent to Microsoft Foundry and set up the workspace and CLI.
- Author an evaluation suite with concrete quality thresholds.
- Add CI/CD release gates and watch them block a deliberate regression.
- Instrument observability for traces, telemetry, alerts, safety events, latency, and cost.
- Connect production traces back to evaluation and continuous improvement.
- Run the full loop on a red-to-green pull request and produce an evidence pack.
Workshop structure
| Block | What attendees do | Built artifact |
|---|---|---|
| Foundations | Deploy travel-agent:1 and set up the agentops-vbd workspace and CLI. | Live agent and configured workspace. |
| Evaluate | Author an eval suite that scores the agent on real criteria. | Passing evaluation suite. |
| Ship | Turn the eval suite into a CI gate that blocks a regression. | Gated, passing pipeline. |
| Observe | Emit traces and correlate runtime telemetry to a release. | Dashboards and correlated traces. |
| Operate | Add safety follow-through and turn a trace into a new eval case. | Standing safety checks and a closed feedback loop. |
| Capstone | Drive a red pull request to green through the full loop. | Production-readiness evidence pack. |
Workshop pages
| Page | Purpose |
|---|---|
| Agenda | Full-day schedule. |
| Lab roadmap | Lab sequence and learning objectives. |
| Observability strategy | Cross-cutting observability model for the full-day workshop. |
| Lab guide | Step-by-step hands-on labs from Lab 1 through the capstone. |
Observability treatment
Observability is both a dedicated lab and a thread across the workshop. Every lab should preserve enough metadata to support release evidence and runtime diagnosis:
- Agent version
- Dataset and eval run
- Release or deployment ID
- Trace ID
- Safety findings
- Owner and incident path
- Runtime telemetry links
What this is not
This VBD track is not an AgentOps Toolkit walkthrough and not an app-build tutorial. It is a production-readiness working session around Microsoft Foundry, repo-side release evidence, CI/CD gates, Azure Monitor and Application Insights telemetry, safety review, and Day-2 operations.