Skip to content

Ask about the evidence.

Understand a finding, summarize its sources or prepare the next review question. AI Advisor uses evidence you are authorized to access, with scope, period and currency.

Example data

Growth and Enterprise include AI Advisor. The conversation here uses synthetic evidence, not customer data or a verified savings result.

AI AdvisorSeptember 2026 · EUR

Why did Production increase?

Production was €56,148.25 in September, €3,962.33 more than August. The largest itemized charge is vm-api-02 at €12,482.36. Its configured owner is Commerce platform. Utilization is not included, so this charge alone cannot tell us whether it is oversized.

  • Azure billing · September 2026
  • Resource inventory
  • Configured mappings

Sample conversation. AI Advisor is included in Growth and Enterprise.

Follow an answer back to its sources.

Scope and grounding
Questions use permitted evidence. Financial context requires an included scope, explicit period and native currency. Missing sources remain limitations.
Provenance
Source basis, policy/prompt context and supported generation metadata remain part of the governed explanation. A response fingerprint is not an atomic database snapshot.
Privacy and authorization
Tenant-scoped authorization and minimized model inputs constrain the source context. The explanation does not grant access or replace server-side permission checks.
Failure behavior
Unconfigured or unavailable generation must be represented honestly. Core non-AI investigation remains usable; fixtures are not customer analysis.

Prepare the owner's next question.

Estimates, assumptions, customer decisions and verified results remain different records. AI cannot certify savings, grant tenant authority or execute customer-cloud remediation.

Review, test and verify

Confirm missing telemetry and workload requirements before testing a change. Your team implements through its own process; provider or billing evidence is needed to verify the result.

AI governance Read-only access and data handling

Start with a question the evidence can answer.

Explore a sample answer, then bring your next cost question to the team.

Compare plans