Skip to main content

Score and route records with AI: a risk or priority score, then an automation

Give records an AI risk or priority score with a Field Agent, then let an automation assign, notify or move them after a person reviews.

Written by Peter Novosel

To score and route records with AI, add an AI Field Agent to a Number field that scores each record from its text. Then build an automation that acts when the score and a review field match a condition. The agent prepares the information and a person approves it before the automation assigns, notifies or moves the record.

Plan Availability

Team, Professional and Enterprise (the plans with SmartSuite AI on the pricing page); the 14-day free trial runs on Professional.

Permissions

Workspace Administrators enable AI Field Agents and manage connections. Solution Managers configure the agent and build the automation. Members run the agent and set the review field.

Platform

Web.

Related Reading

AI Field Agents: configure, run and troubleshoot
​Automations Introduction
​Automation Conditions
​Automation Action: Conditional Logic.

Step 1: score each record with an AI Field Agent

  1. Add a Number field, for example Risk score.

  2. Open its Field Settings, select the AI Field Agent tab and turn on Enable AI Field Agent.

  3. Write the prompt with the scale and the criteria, and insert the fields it depends on, for example: Score the risk in {{Risk Description}} from 1 to 100. Likelihood counts for half and {{Impact}} for half. 90 or more means an immediate threat to operations; 50 to 89 means a loss that needs a mitigation plan; under 50 means routine. Return the number only.

  4. Choose a connection and turn on Run Automatically.

  5. Click Update Field, then run the agent on a few records and compare the scores with your own judgment. Adjust the criteria in the prompt until they agree.

Step 2: add a review step

Add a Single Select field such as Review with the options Pending, Approved and Rejected, defaulting to Pending. The score is the agent's suggestion; the person who owns the record sets Review to Approved after reading it. The automation in the next step waits for that value, so nothing is assigned or escalated on the score alone.

Step 3: route with an automation

  1. Open Automations from the Solution Menu and click Add Automation.

  2. Choose the trigger When a record is updated or When a record matches a condition, and set the condition: Review is Approved and Risk score is greater than or equal to 80.

  3. Add the actions: Update a record to set the Assigned to field and a Status such as Escalated, and Send a notification or Send an email to the owner.

  4. Optional: add Conditional Logic to branch, for example one branch for scores of 90 and above that notifies a manager, and another for 80 to 89 that assigns a reviewer.

  5. Name the automation, turn it on with the toggle, and test by approving a scored record.

Note: each action in a run counts toward your plan's automation actions per month, and a record that already matched the condition when you turned the automation on does not trigger it until it changes.

How do I score inside the automation instead?

When the score is only needed for routing, replace the Field Agent with an AI Workflow Agent step in the automation. Give it a Custom output with a Number value for the score and a Single Select value for the band (High, Medium, Low). Then use Conditional Logic on those outputs and an Update Record action to store them. The review field still gates the actions.

Frequently asked questions

Can the automation act before anyone reviews the score?

Only if you let it. Put the review field in the trigger condition, as above, and the automation waits for Approved.

Does the score update when the record changes?

Yes, when Run Automatically is on: the agent re-runs whenever a referenced field changes, and the activity history shows each run as Updated by AI.

Why do two similar records get different scores?

The model interprets the prompt. Give it a scale, explicit criteria and the fields that matter, and keep the prompt the same across records; test on known cases before relying on it.

Can I see why the agent chose a score?

Ask for it. Add a second agent on a Text Area field, for example Score rationale, with a prompt that explains the score in two sentences from the same fields.

Last verified 24 Oct 2026.

Did this answer your question?

Related Articles