Select any Flow and navigate to the Integrations tab and then select the Agent Tool Scoring tab

AI agent tooling has historically centered almost exclusively on API-based tools. As a result, using AI agents to automate interactions with legacy systems where APIs aren't available has been difficult.

By using the AI agent as the orchestration layer and exposing your existing RPA flows (or the components within them) as the available tools, agents can automate tasks in systems that have no readily available API — backed by automation that has already been vetted and relied upon in production.

The Agent Tool Scoring page analyzes and grades each PAD component in your currently selected EstimALC Environment to identify the strongest candidates to become agent tools.



Blueprint looks for 3 types of components:

Each component receives a single grade, shown as a percentage, that reflects how ready it is to become an agent tool. Higher scores indicate components that are easier and safer to expose as a tool — generally smaller, self-contained units with explicit inputs, limited branching, and few hidden dependencies.

Click any score to open a side panel with the full breakdown of the individual metrics that contributed to the grade. The factors considered include the number of actions, the number of arguments, unique applications touched, decision complexity (Ifs, Switches, and Loops), undeclared variables, exception handling, and any hard-coded sensitive values. Not every factor is weighted equally; the metrics that most affect a component's suitability as a tool contribute the most to the overall grade.

Most of a component's score is static and are recalculated only when you modify the underlying process and re-import it. The exception is runtime data: when Include Runtime Data is enabled, Blueprint updates those metrics on a semi-real-time basis as executions happen, so the portion of the score derived from runtime data can change without a re-import.