Skip to content
FDAI Forward Deployed Agents

Cost Governance

FDAI watches your cloud spend the way it watches everything else. Detection is deterministic first, autonomous action covers only the low-risk majority, and anything that could hurt waits for human approval. FDAI finds waste, proposes right-sizing, and cleans up the safe subset on its own. Changes with a real impact scope wait for you.

What you get

Section titled “What you get”
  • Spend anomaly detection. A cost signal that drifts from the expected baseline raises a detected issue. Detection runs in observation mode and never acts on its own.
  • Right-sizing recommendations. FDAI flags an over-provisioned resource with a concrete, reversible fix.
  • Safe cleanup, automatically. The low-risk subset, such as idle disk cleanup, unused public IP release, and orphan NIC removal, runs on its own with a rollback path.
  • Risky cost changes wait for you. Anything above the safe threshold goes to human approval and is never applied automatically.

How agents govern cost

Section titled “How agents govern cost”

Njord owns cost anomalies and recommendations but remains advisory. Forseti judges the proposed response, while Odin arbitrates cost against reliability and capacity objectives. Var owns required human approval, Thor alone executes, Heimdall independently verifies realized utilization and cost effects, Vidar controls recovery, and Saga records the outcome. Cost evidence never grants itself execution authority.

How a cost action reaches enforcement

Section titled “How a cost action reaches enforcement”
  1. Detect the anomaly. A cost-anomaly detector fires on, say, an over-provisioned cache tier and raises a normalized detected issue.
  2. Match a rule. The deterministic tier (T0) matches the detected issue to a right-sizing or cleanup rule.
  3. Prove it in observation mode. The rule runs in observation mode, judging and logging without changing anything, until it clears its promotion gate.
  4. Promote to enforcement. The action becomes autonomous only after the measured accuracy holds up.
  5. Ship with a rollback. The right-size or cleanup lands as a fix pull request that carries its own rollback reference and audit entry.

Proof, not promises

Section titled “Proof, not promises”

Cost governance is measured against a baseline, never asserted (see goals and metrics and the illustrative cost model):

  • Cost per unit, reported as $/optimization for cost actions, is a target to lower. FDAI states it only once the baseline and the treatment have been measured on the same scenario set.
  • Rollback rate is a guard metric and should not rise above the baseline.
  • FDAI never claims a cost multiplier without a paired measurement.
Section titled “Related”
Deterministic firstWhy detection stays rule-driven and reviewable.Risk tiersHow a cost change becomes auto, human approval, or deny.Observe, then enable changesHow a cost action earns autonomy.Cost modelThe illustrative Azure cost envelope.Deploy and onboardBring FDAI into your environment.