1) Where observability becomes “decisioning,” not just dashboards

• Where is AI delivering the “first wave” of value in ops (correlation, anomaly detection, RCA), and where do humans stay essential?
• Which signals are actually feeding AI today (telemetry, events, tickets, runbooks) and what’s still missing for reliability-grade outcomes?

2) Scaling agentic AI: the operating model shift (and what blocks it)

• What’s the hardest barrier to moving from pilots to scaled production—data quality, process change, risk appetite, or executive sponsorship?
• What changes when agents can take action (approval tiers, rollback safeguards, accountability, new roles)?

3) Trust, security, and governance: guardrails that unlock autonomy

• What threat model matters most when agents operate across IT workflows (e.g., prompt/runbook injection, privilege escalation, drift, telemetry poisoning)?
• What “guardrails” are proving practical across industries: policy constraints, audit trails, kill switches, staged rollouts, and least-privilege identity for agents?