Every FinOps team has questions no off-the-shelf tool can answer. The workflows that matter most are specific to your cloud, your org, your owners, and your operating model.
This session shows how to build them yourself. An AI coding tool plus the Oracle Cloud MCP (Model Context Protocol) gives practitioners everything they need to build personal, governed, action-oriented agents around the questions they actually ask.
You’ll see four agents built on stage, each more capable than the last:
-
A waste scan that identifies idle and oversized resources across your tenancy, ranked by monthly impact and owner
-
An investigator that explains why each finding is wasteful, who owns it, and how confident the agent is in the recommendation
-
An always-on agent that watches your tenancy continuously and surfaces only what’s new, growing, or risky
-
An action loop that turns each recommendation into an approved, applied Terraform change
We’ll close with where this leads next: agent memory architecture that learn what real workloads looks like, multi-agent orchestration that lets specialized agents hand off work, and a multi-cloud agent fabric where the same pattern works across every cloud your org runs.
You’ll leave with the recipe, the guardrails, and the prompts to try out first thing Monday morning.
