Set Up Policy Agent

Overview

Available on Ramp Plus. See Ramp Plus overview for plan details.

Policy Agent is Ramp’s AI expense reviewer that applies your written expense policy to every card transaction and reimbursement. This article walks through how to onboard Policy Agent, convert your existing policy into an AI-friendly structure, and publish the agent-facing policy so it can start evaluating expenses.

Policy Agent starts in review-only mode — you decide when to enable automations later.

Access and prerequisites

To set up Policy Agent, you must:

• Be on Ramp Plus
• Be an admin with access to the Policy tab

In the Ramp app, go to:

• Policy

From there, you can start configuring Policy Agent’s knowledge center.

Choose how to start your policy

In the Policy tab, you’ll see options to onboard Policy Agent using your existing materials:

• Use Ramp’s best-practice template
  • Good if you don’t have a detailed expense policy yet.
• Use your existing employee-facing policy as a starting point
  • Good if you already have a policy in your handbook or PDF and want the agent to mirror it.
• Upload another policy document
  • Good if your expense policy lives in a separate doc you maintain elsewhere.

Regardless of which path you choose, Ramp converts your input into an AI-friendly structured policy that the agent can interpret and cite.

FAQ – How does my existing policy get formatted for the agent? Ramp transforms your long, human-written expense policy into a concise, structured format that Policy Agent can understand best. This makes Retrieval-Augmented Generation (RAG) more effective, so the agent can look up and reference the correct part of the policy in real time when analyzing every expense.

Document size and format

To get the best results from Policy Agent, prepare your policy document with these constraints in mind:

• Recommended limit: ≈10 pages / 10,000 words
• The agent cannot interpret images or hyperlinks
  • Any rule that matters for enforcement must be written in plain text
• Avoid unnecessary filler or narrative content that doesn’t drive a decision

FAQ – Is there a certain page limit or file size for policy upload? Approximately 10 pages / 10,000 words is recommended.

FAQ – How does the agent handle images or hyperlinks? The agent cannot interpret images or hyperlinks, so make sure all important context is explicitly stated in text.

How conversion to AI-friendly format works

After you select a starting document and kick off conversion, Ramp:

• Parses your existing policy
• Strips filler and non-enforcement text
• Restructures content to highlight clear, enforceable rules
• Organizes the policy into a concise, structured format optimized for AI understanding and citations

This is what allows Policy Agent to:

• Retrieve the right rule at evaluation time
• Cite the exact policy text it used for each recommendation

What changes (and what doesn’t)

During conversion:

• Content that is not helpful for policy guidance or enforcement may be removed
• No new policy rules are added
• Images and hyperlinks are ignored as a source of policy logic

Your underlying policy intent stays the same — it’s just presented in a structure the agent can interpret reliably.

FAQ – What happens when the policy agent is making my existing policy AI friendly? The conversion strips filler and non-enforcement text and restructures content for clarity. It does not add new rules. The result is a concise, structured version of your existing policy optimized for the agent’s understanding and citations.

FAQ – Will any content be removed or added from my existing policy when I convert it to an AI-friendly format? Any content that is not helpful in policy guidance or enforcement will be removed. No new policy content will be added.

Conversion status and notifications

Once conversion starts:

• The process runs asynchronously (you can safely leave the page)
• You’ll see progress states (e.g., “Request received”, “Reading doc”)
• You’ll be notified on your homepage when conversion is complete

FAQ – If I exit the page during conversion, will my progress be lost? No. Once the process is kicked off to make the policy AI friendly, you will not lose your progress by leaving the page. Your progress will be trackable, and you’ll be notified on the homepage once it’s complete.

Publish the agent-facing policy

Once conversion is complete, you’ll land in the Policy Agent editor with a draft version of your agent-facing policy.

In this editor, you can:

• Review the structured rules and sections
• Clarify ambiguous language
• Add per-entity, per-department, or per-location rules in plain English
• Add hidden notes for sensitive exceptions (visible only to admins and the agent)

When you’re satisfied with the draft:

1. Review each section for accuracy and completeness
2. Confirm that all important rules appear in plain text (not only in images or links)
3. Click Publish to make the agent-facing policy live

Publishing:

• Updates the agent’s knowledge center
• Does not change the employee-facing PDF by itself
• Does not force employees to re-sign the policy

Initial evaluation timing

After you save and publish the agent’s knowledge:

• Policy Agent evaluates recent expenses against your policy
• This initial evaluation can take up to 24 hours
• You’ll be notified on your homepage when this backfill is complete

Once the initial evaluation finishes:

• New card transactions and reimbursements are evaluated as they come in
• Expenses will start appearing in groups like Approval recommended and Review recommended
• Policy Agent remains in review-only mode until you explicitly add it to approval workflows

Agent-facing vs Employee-facing documents

As part of setup, it’s important to understand the two distinct policy documents you’ll manage in the Policy tab:

Policy Agent document / knowledge center (Admin-facing)

• Structured, AI-optimized policy that Policy Agent uses to evaluate expenses
• Contains rules, hidden notes, and references to custom fields, entities, and programs
• Feeds the agent’s reasoning and citations
• Editable at any time without directly changing what employees see

Employee-facing policy document

• The PDF or handbook version that employees see and optionally sign
• Not automatically updated when you edit the agent-facing document
• Can be replaced at any time via Policy → Documentation

FAQ – Does it overwrite the existing PDF policy? No. Your existing employee-facing PDF policy remains unchanged. Policy Agent uses a separate agent-facing knowledge center.

FAQ – Do changes re-trigger employee sign-off like the original PDF? No. Modifications to the agent-facing policy do not force all employees to re-sign. The employee-facing policy document remains separate. If you want employees to sign again, you can save the updated policy document as a PDF and re-upload via Policy > Documentation.

FAQ – Is there a way to toggle whether the refined AI policy can overwrite the existing PDF? No. There isn’t a toggle for the agent document to automatically replace your employee PDF. You can download or export a structured version and upload a new PDF if you’d like them to align.

FAQ – How can we see which PDF was uploaded for the policy agent? Not currently supported. You can replace the employee-facing PDF at any time via Policy → Documentation.

FAQ – How are employees notified of policy changes to Policy Agent’s knowledge center? They aren’t. The agent’s knowledge center is separate from the employee-facing PDF and signature experience. Changes you make there don’t trigger notifications or new acknowledgments.

Policy page overview

The Policy page provides a high-level view of:

• The policy knowledge that Policy Agent is using (distinct from the employee-facing policy)
• How often the agent is recommending approvals or rejections vs. being unsure
• Alignment (how often reviewers agree with Policy Agent)
• Quick links to visibility settings (who can see assessments)

This page is your main control center for Policy Agent configuration and monitoring.

Testing with real transactions

Policy Agent runs only in production, but you can safely test it before rolling it out broadly.

FAQ – Is there a way to test Policy Agent with real or sample transactions before going live? Policy Agent only runs in production. After setup, visibility defaults to All reviewers, which is also the recommended setting for most teams. If you want to test privately first, you can temporarily switch visibility to Admins only while you refine your policy configuration.

Recommended testing approach:

1. Publish your agent-facing policy
2. Confirm visibility is set to All reviewers in Policy → Expense reviews (this is the default after setup)
3. Review the agent’s assessments and citations on real expenses
4. Refine policy language if needed
5. If you want to test privately first, you can temporarily restrict visibility to Admins only . When you’re ready, set visibility back to All reviewers and (optionally) add Policy Agent into approval workflows