SAP BTP Application and AI Consumption Monitoring

Steps

1. Overview
2. Pre-requisites
3. Key Choices and Guidelines
4. Implementation

1 Overview

Description

Enterprise teams can operate dozens of SAP BAIP applications, APIs, agents, and published demos. The information needed to operate them is usually distributed across a portfolio catalog, Cloud Foundry, logging, event services, and consumption reporting.

This Best Practice describes a read-only monitoring pattern for SAP BAIP. It combines approved portfolio data with current Cloud Foundry state, operational evidence, lifecycle events, application-defined LLM telemetry, and official SAP BAIP usage data where available. Each source answers a different operational question. The report shows the source of the evidence rather than treating every signal as proof of the same condition.

For agents, a running process is only one part of the picture. It does not show whether an agent used approved tools, stayed within its assigned scope, or achieved the intended business result. This Best Practice covers monitoring evidence, not the full set of controls needed to govern an agent. Action approval, business outcomes, and risk decisions need separate owners and data. For the wider governance context, see Marco Cigaina's Agentic AI Governance.

Expected Outcome

After implementing this pattern, a team can:

• define a monitored application scope from a maintained inventory;
• verify whether a mapped Cloud Foundry application has running instances;
• investigate recent request volume, 5xx responses, latency, error signals, and crash signals;
• retain lifecycle events separately from log-derived observations;
• analyse model calls, tokens, cached tokens, errors, and latency by application, model, and calendar week;
• compare operational AI telemetry with official account-level consumption at its reported granularity.