AI Features Overview#
Reply CMP includes an AI layer that turns cloud data into answers, insights, and recommendations — without requiring you to write queries, export spreadsheets, or read native cloud advisor portals.
Five AI capabilities are available across the platform:
The five capabilities#
CMP AgentA conversational assistant embedded in the platform. Ask questions about your cloud resources, costs, and recommendations in plain English. Request on-demand HTML reports.
Example questions:
“Which connections are spending the most on compute this month?”
“Generate a cloud cost report for last quarter”
“How many untagged Azure resources do I have?”
“Show the relationship graph for this database”
AI-enhanced cost savings estimates on cloud provider recommendations. Goes beyond native recommendation tools by computing realistic monthly savings estimates with confidence ratings, capped to actual resource costs.
An adaptive model forecasts each resource’s spend over the next 30 days. The model updates continuously as new billing data arrives — no manual retraining or configuration required.
Create and schedule cloud cost reports through a conversational experience embedded in the Reports section. Describe what you need, preview the report, save the design, and refresh it with new data on every scheduled delivery.
When a resource changes unexpectedly (detected in change history), the AI suggests what CLI commands or Terraform imports you would need to restore the previous configuration.
Data Privacy#
Important
What data does the AI see?
The Reply CMP AI components access the following data:
Resource metadata (names, types, regions, tags) from your discovery inventory
Discovery enrichment data such as support level, type-specific properties, and resource relationships
Cost and billing data from your cloud connections
Allocation group structure (names and hierarchy)
Recommendation data from cloud provider advisors
The AI components do not access:
Raw application data inside your cloud resources (databases, blob contents, etc.)
Secrets or credentials stored in Azure Key Vault
Personal user data beyond display names used for attribution
All AI processing uses Azure OpenAI in your tenant region. Data does not leave your Azure tenant boundary.
Capability Limitations#
Capability |
Limitation |
|---|---|
CMP Agent |
Operates on data up to T-1 (yesterday). Real-time data is not available. |
CMP Agent |
Cannot modify resources (except triggering a Discovery sync, which requires confirmation). |
Savings Estimation |
Estimates are directional. Low-confidence estimates have no cost history to base on. |
Cost Forecasting |
Point estimates only (no confidence interval). Accuracy is lower for highly volatile resources. Requires 7 days of history to activate. |
AI Reports |
Report design sessions expire after 30 minutes. The AI uses yesterday’s data for previews. |
Rollback Suggestions |
Suggestions are advisory only. Apply with caution in production. |
Next Steps#
Chat with your cloud data in natural language.
Learn how confidence ratings and caps work.
How the adaptive online model works and how to read forecasts.
Design and schedule reports through a conversational AI session.
How reusable scheduled report designs work.
Understand the advisory suggestions available in Discovery history.