Summary:
Rovo Agent invocations within Jira Automation are currently governed by a fixed per-tenant service limit. In high-scale environments where automation rules are triggered in large bursts, tenants can exceed these limits, resulting in non-deterministic "Agent invocation failed" errors.

Currently, these failures do not provide specific feedback in the Automation Audit Logs indicating that a rate limit was hit. Furthermore, there is no built-in mechanism to queue or retry these requests, leading to the immediate failure of critical downstream automation steps when the AI Gateway returns a 429 (Too Many Requests) response.

Suggestion:

1. Scalable Throughput: Implement tiered or configurable per-tenant invocation limits to accommodate high-volume automation workloads that require high concurrency.

1. Detailed Audit Logging: Replace generic "Agent invocation failed" messages with specific error details when a rate limit is reached (e.g., "Rate limit exceeded: 429"). This allows admins to identify if the failure is due to a prompt error or infrastructure constraints.

1. Automated Retry Mechanism: Introduce native retry logic with exponential backoff for the "Use Agent" action specifically for rate-limiting scenarios, ensuring temporary bursts do not cause permanent rule failures.

1. Execution Throttling: Provide administrative controls to stagger or throttle the execution of automation rules that utilize AI Agents to help stay within platform guardrails.