The Agent Finder feature intelligently scans your UPN business process diagrams to pinpoint where AI agents, AI workflows, or deterministic automation can be applied. It provides tailored recommendations grounded in a rigorous methodology, helping you unlock greater value from your Salesforce implementation—smartly and efficiently.
Prerequisites
You need requirements manager permission on the space
To use Agent Finder, you need to have it enabled as part of your license. It is part of Diagram Insights add-on.
When enabled: You can run reviews and receive actionable AI recommendations.
When disabled: The tab remains visible but shows a placeholder message:
Agent Finder is not available on your license. Learn more about your journey with Elements.cloud.
This flag is enabled by default for users on consulting license.
Agent Finder works best on diagrams that follow Universal Process Notation. It is advisable that before running Agent Finder, you automatically review your process diagram using our Standards Check.
How to launch Agent Finder
To access the feature, open any diagram of type UPN:
Switch to Edit mode
Navigate to the Insights tab on the right panel.
Select the Agent Finder sub-tab.
Click Find Agent Opportunities to initiate a new AI-powered review.
This tab is exclusive to UPN diagrams and is not accessible for other diagram types.
What Agent Finder Does
Agent Finder applies a structured methodology to analyze each activity in your UPN diagram. It identifies the most effective type of automation—Conversational Agent, AI Workflow, or Deterministic Automation—based on how the task is triggered, the type of input it requires, the nature of the activity, and the expected outcome.
Conversational Agent
When to use:
For tasks that are judgment-based, variable in nature, and benefit from interactive, adaptive handling.
Trigger/Input:
Initiated by unpredictable, situational events like user chats or handovers.
Inputs often include fuzzy or unstructured data such as messages, notes, documents, or recordings.
Activity Characteristics:
Requires probabilistic reasoning or interpretation.
Common verbs include: analyze, generate, document, summarize, propose.
Expected Output:
Produces contextual or judgment-based results like a business proposal, a tailored recommendation, or a summarized plan.
AI Workflow
When to use:
For record-triggered tasks requiring AI interpretation but with a clear, structured follow-up.
Trigger/Input:
Triggered by specific events like record creation or updates.
Inputs are often unstructured but interpretable, like documents or emails.
Activity Characteristics:
Involves AI-enabled analysis or transformation with predictable logic.
Often uses verbs like classify, extract, identify, write.
Expected Output:
Leads to a discrete, actionable result such as a new record, updated field, or a database operation.
Deterministic Automation
When to use:
For routine, rules-based operations that rely on structured data and fixed outcomes.
Trigger/Input:
Starts with clearly defined, structured conditions (e.g., field value changes, record status).
Inputs are precise and concrete, such as dates, numerical values, or status fields.
Activity Characteristics:
Executes deterministic logic with no need for interpretation.
Verbs typically include calculate, update, assign, notify, validate.
Expected Output:
Generates precise and consistent results, like updating a record, sending a notification, or assigning a task.
Each recommendation is presented with:
A confidence level (minimum 80% is required for recommendation to be generated),
A justification explaining triggers, inputs, reasoning, and expected outcome,
Suggested resources like an AI Agent, Prompt Template, or Flow.
How to View Recommendations
After triggering a review:
A notification appears when processing is complete.
If you're already on the Agent Finder tab, results automatically display—no need to navigate away and return.
Recommendations are shown in a structured list under the Insights > Agent Finder tab:
Each includes a summary of the related parent activity
Hovering reveals full before-and-after context for clarity.
The associated diagram activity is highlighted when you hover over a recommendation.
Taking Action on a Recommendation
When you accept a recommendation:
The proposed resources (AI Agent, Flow and Prompt Template) are added to the diagram activity. Existing resources remain unchanged.
A business requirement is created with:
Summary:
AI: [Activity Text](e.g., “AI: Document the customer’s goals and desired outcomes”)What is required: Explains the need (e.g., “We need to accelerate and improve step ‘Document the customer's goals’ in the 'Customer onboarding’ process.”)
How it might be implemented: A direct quote of the rationale from the recommendation.