Skip to main content

AI Prompt

AI Prompt is a workflow trigger type that lets you describe an approval condition in plain English instead of picking from a fixed list of deal attributes. When a deal is submitted, RevOps sends your statement along with the deal's terms and line items to an AI model, which decides whether the statement holds true for that deal.

AI Prompt triggers are only available for organizations on the RevOps enterprise plan.

How it works​

Every AI Prompt clause is a single statement describing something that should be true about a deal, for example: "Any line item is discounted more than 30% off its list price."

When a deal is submitted for approval, RevOps sends each statement along with the deal's terms, line items, and other details to an AI model. The model returns:

  • Whether the statement is true or false for that deal
  • A supporting quote from the deal
  • A short explanation of its reasoning

That true/false result is treated exactly like any other clause result — it can trigger auto-approval, assign an approval topic, or assign a countersigner, depending on how the workflow is configured. See Creating Workflows for how triggers fit into the rest of a workflow.

What data the AI can see​

When a deal is submitted, the model receives:

  • Deal terms — every term configured on the deal (for example Payment Terms, Contract Length, Auto-Renewal, and any custom terms your organization has added), including each term's name, description, and its selected value for that deal.
  • Line items — every SKU/line item on the deal's quote: its name and description, list price, adjusted (negotiated) price, whether it's flagged as a discount or credit, the discount percentage, and its pricing schedule.
  • Deal-level details — the agreement template used, the deal's creator and owner (the internal users associated with the deal), and the quote subtotal.
Not included

The model does not receive the customer's signer/contact details, account or CRM data (such as industry or company size), or any of your organization's other historical deals. Statements that depend on this kind of information can't be evaluated reliably — see Writing effective prompts below.

Configuring an AI Prompt trigger​

  1. Go to Settings > Workflows and open (or create) a workflow.
  2. Under Workflow Triggers, add a new clause and select the AI Prompt tab.
  3. Enter your statement in the Value field. Statements are limited to 1000 characters.
  4. Save the workflow.

Writing effective prompts​

Since these statements are evaluated by an AI model rather than matched against exact field values, keep them:

  • Specific and unambiguous — reference concrete deal attributes (discount percentage, term length, payment terms) rather than vague business judgment calls.
  • Self-contained — the model only sees the current deal. See What data the AI can see for exactly what that includes.
  • Focused on one condition per statement — combine multiple AI Prompt clauses rather than writing a single compound statement.

Example prompts​

These examples reference data that's directly present on every deal — pricing, terms, and line items — which gives the AI model the clearest signal to evaluate against:

  • "Any line item is discounted more than 30% off its list price"
  • "The total contract value is more than $10,000 below list pricing"
  • "The selected Payment Terms are anything other than Net 30"
  • "The contract length is longer than 3 years"
  • "The deal includes an add-on SKU without its required base product SKU"

Test any prompt against a few real deals before relying on it — statements that reference information outside the deal itself are harder for the model to evaluate reliably, since that data isn't part of what's sent for evaluation. See What data the AI can see for the full list of what's included.

Debugging​

Like any other trigger, you can see how an AI Prompt clause evaluated for a specific deal from the workflow results section on the deal overview page. The result includes the AI's true/false answer along with the supporting quote and reasoning it returned, so you can confirm why a deal did or didn't match.