Challenges — AgentOps Control Tower
Welcome to the build. These seven challenges take you from an empty subscription to a fully operational AgentOps Control Tower on Microsoft Fabric.
How the challenges work
- Goal-oriented, not click-by-click. Each challenge gives you a scenario, clear objectives,
and success criteria. How you get there is your team's call. The
resources/folder is your toolbox. - Sequential. Each challenge builds on the previous one — complete them in order.
- Checkpoints. When you meet the success criteria, show your coach for a ✅ before moving on.
- Hints are there if you need them. Each challenge ends with hints and links. Try first, then peek. Coaches hold the full reference solution.
The arc
| # | Challenge | Depends on | Est. | Theme |
|---|---|---|---|---|
| 0 | Mission Briefing & Environment Setup | — | 1–1.5 h | Access, tooling, Fabric capacity |
| 1 | Light Up the Agents | 0 | 1.5–2 h | Deploy a Foundry agent workload that emits telemetry |
| 2 | Build the Telemetry Landing Zone | 0 | 1.5–2 h | Land cost, metrics, logs & metadata in ADLS Gen2 |
| 3 | Connect Fabric to the Enterprise | 1, 2 | 2 h | Lakehouse, OneLake shortcuts, Cosmos DB Mirroring |
| 4 | Refine the Signal | 3 | 2–2.5 h | Bronze → Silver → Gold medallion + pipeline |
| 5 | Stand Up the Control Tower | 4 | 2–2.5 h | Direct Lake model + Power BI dashboards |
| 6 | Make It Operational 🏆 | 5 | 1.5–2 h | Stretch: alerts, chargeback, RLS, automation |
💡 Short on time? Challenges 1 and 2 are independent of each other and can be tackled in parallel by splitting your team. They must both finish before Challenge 3.
Scoring
This is a learning RVAS, not a competition — but if your event wants friendly scoring, here's a simple rubric (coaches can adapt). Points are awarded per challenge at the success checkpoint.
| Challenge | Base points | Bonus opportunities |
|---|---|---|
| 0 | 5 | Everyone on the team can az login + reach Fabric (+5) |
| 1 | 15 | Custom token/cost metric visible in App Insights (+5); end-to-end trace shown (+5) |
| 2 | 15 | All 5 data domains landing (+5); FOCUS cost file validated (+5) |
| 3 | 20 | Mirroring latency < 1 min demonstrated (+5); zero-copy explained to coach (+5) |
| 4 | 25 | Gold correlates cost and agent telemetry (+10); pipeline runs green end-to-end (+5) |
| 5 | 20 | All three dashboard themes built (+5); Direct Lake confirmed (+5) |
| 6 🏆 | 20 | Live alert fires (+5); chargeback by team/use case (+5); RLS (+5) |
Judging themes (for a showcase at the end): Does the Control Tower answer the three business questions — reliability, cost, performance? Is the data correlated, not just collected? Could the customer actually run on this?
What "done" looks like
By the end of Challenge 5 your team can stand in front of the customer's leadership and answer, live:
- Which agents are healthy right now, and where are the errors? (Reliability)
- What is each agent / team / use case costing us? (Cost)
- Are we meeting our SLAs, and how much are we scaling? (Performance)
Challenge 6 makes it operational — the Control Tower starts telling you when something's wrong.
Ready? Start with Challenge 0 — Mission Briefing & Environment Setup.