Discovery Found Everything. Your Agent Still Doesn't Know What Matters
I have an AI service desk analyst agent running in production right now. It completes work on its own, without a person confirming it. It is also restricted to one incident type on one service, and this session is about why.
The agentic testing conversation has focused on one question. Is the agent reasoning correctly? We build evaluation suites, we score outputs, we trace decisions. What we have not examined is the quality of the enterprise context the agent reasons over, and the assumption underneath it that a populated CMDB means an informed agent.
That assumption deserves scrutiny, but not the kind it usually gets. Where automated discovery is running, the asset layer is often excellent. Discovery runs on a schedule and finds what is actually deployed. The CMDB is accurate about what exists. It is far less reliable about what those things mean to the business. Service mapping is a project rather than a scheduled job, and most organizations never finish it. Criticality was entered once at build time. Ownership drifts after every reorganization. Failure behaviour is discovered during outages and rarely written down. Every layer carrying meaning is human-maintained, and nothing forces it to stay correct.
This was survivable while humans read the CMDB, because experienced people silently correct it as they go. An agent cannot. It reads the record and accepts it exactly as written. We chose our first use case the way everybody does, on ticket volume, and it was the right decision. But volume told us how often the work happens. It told us nothing about whether our data could support an agent doing it. We got a good outcome by coincidence, and the next use case is where that stops working.
Drawing on two decades building and running configuration management, change governance and testing in an organization where a failed change is a public service disruption, this session offers a four-part Context Integrity Test that fits inside the change process you already run.
Key Takeaways:
Why an accurate CMDB and a trustworthy CMDB are different things, and why discovery solves the first and not the second.
The four layers that carry meaning rather than inventory, and how each one degrades.
Why choosing an agent's first use case on volume works, and why it fails for the second one.
A four-part Context Integrity Test, and how to match an agent's permission level to the data quality of the specific service it touches.
About the speaker
Angela Cutulenco:
Angela Cutulenco is Senior Director of Governance, Administration and Continuous Improvement in the office of the CIO at the Toronto Transit Commission, one of North America's largest public transit systems. She has 33 years in IT and over 30 years in progressive leadership roles across quality management, IT service management, testing, and governance. She designed and implemented the TTC's first configuration management database and used it as the basis for change impact assessment, and spent six years as Supervisor of Change, Configuration and Release Management, chairing the weekly Change Advisory Board. She is Product Owner and subject matter expert for the TTC's integrated ServiceNow platform, spanning IT Service Management, IT Operations Management, IT Asset Management, Service Mapping, hardware and software asset management, automated discovery, and Strategic Portfolio Management. She is currently leading the deployment of an AI service desk analyst agent, which runs in parallel across non-production and production. In parallel she has led end-to-end testing and quality assurance across the TTC's SAP portfolio, including ECC, S/4HANA, SuccessFactors and Time and Attendance, delivering major production transitions with zero Priority 1 incidents and under two percent defect leakage. She began her career in international consulting with Price Waterhouse, Booz-Allen & Hamilton and KPMG in Moldova, and later implemented configuration management at TNK-BP in Moscow, the first implementation of its kind in Russia. She is an ITIL Expert and ITIL 4 Leader in Digital and IT Strategy, with additional training in COBIT, TOGAF, and information systems auditing. She holds an Executive MBA from the Ivey Business School at Western University.
About
TestMu Conf
Testμ (TestMu) is the world’s largest virtual conference on agentic engineering and quality, built by the community, for the community. As AI reshapes how we build, test, and ship software, Testμ Conf is where you connect, grow, and lead: agentic workflows, autonomous quality, battle-tested AI playbooks, hands-on workshops, and the engineering culture driving it all.
More Sessions
Join the builders, testers, and innovators shaping the next generation of web experiences.
Testμ Conf 2026 is where they meet.
