How to set up knowledge base for payment plan inquiry chat

You need a Chatref account (free $50 credit, no credit card required) and your payment plan documents: written policies, accepted payment methods, installment terms, late-fee rules, and any eligibility criteria. Have these ready as PDFs, web page URLs, or plain text.

If you handle Medical Billing Services, this setup anchors answers in your practice’s exact billing policies, not generic guesswork. Nothing else is required – no coding, no servers.

1. Create your agent
   Log in to your Chatref workspace, click New Agent, and give it a name (e.g., “Payment Plans”). Keep the default prompt for now; you will tune it later.

2. Upload your payment plan content
   Open the agent’s Knowledge Base tab. Drag your PDFs, paste the URLs of your billing pages, or type key policy snippets directly. Include details like:

   • Payment plan types (monthly installments, hardship plans, interest-free options)
   • Amount thresholds or service-specific plans
   • Enrollment steps (forms, approvals)
   • Late-fee and default consequences Chatref reads everything you add and grounds its answers only in that content.

3. Set up a custom action (optional but recommended)
   Go to the Custom Actions tab. Create an action named “Payment Plan Inquiry.” Define the fields you want to collect in-chat, for example:

   • Patient name
   • Account or invoice number
   • Preferred plan type
   • Preferred payment date
   • Contact preference (email, phone)
     This turns the chat from a simple FAQ into a live intake form. Later you can configure a webhook to send the collected data straight to your billing system (see the custom actions guide for the trigger setup).

4. Link the custom action to the agent
   After saving the action, attach it to your payment plan agent. Edit the agent’s prompt to mention the action, so when a patient asks about payment plans, the agent knows to offer the intake form:
   “When a patient asks about setting up a payment plan, use the Payment Plan Inquiry action to collect their details before providing plan options.”

5. Refine the prompt
   Still in the agent settings, add a short instruction that matches your billing voice, e.g., “Always answer payment questions from the uploaded documents. If the document doesn’t cover a specific scenario, say ‘I can collect your details and have a billing specialist follow up.’”

Use the live Playground on the left panel. Try these queries and verify the answers:

• “What payment plans do you offer?”
• “Can I pay in installments for a $2,000 procedure?”
• “What happens if I miss a payment?”
• “How do I enroll for the hardship plan?”

Confirm that every response is pulled from your uploaded documents – no invented terms or numbers. Then test the custom action: ask “I’d like to start a payment plan,” and ensure the chat prompts for the intake fields and captures them.

Also try a question your docs don’t cover, like “Can I pay with cryptocurrency?” The agent should politely decline and offer to hand off to a person, not make up an answer.

Answers are too vague or off-topic
Your uploaded documents may not be detailed enough. Add concrete terms: dollar amounts, plan duration, eligibility rules, and exact wording from your legal agreements. Broad policy statements lead to generic bot replies.

Custom action doesn’t trigger
Check that the action is assigned to the correct agent and that your prompt instructs the agent when to use it. The phrase that triggers it – “set up a payment plan,” “payment plan,” “installments” – should match what patients naturally type. Edit the agent’s prompt to include those trigger phrases explicitly.

Bot answers with information that isn’t in your docs
This rarely happens with a properly grounded agent, but if it does, tighten the agent’s system prompt with: “Never invent details. If the answer isn’t in the provided documents, say so and suggest a human follow-up.” Then re-test with the same query.

The conversation stalls after collecting details
If you haven’t connected the custom action to a webhook yet, the collected data sits in the conversation log. To close the loop, set up a webhook that sends the intake data to your billing system or internal ticketing tool. Without it, you will need to manually review the logs and email the patient separately.

Patients ask about their specific account balance
The knowledge base alone cannot answer account-specific questions because you have not uploaded individual patient data (and you should not). That is by design. The custom action can capture the patient’s details so a billing specialist can look up the account later, or you can connect a read-only database via a webhook if supported by your EHR/billing platform.

What causes payment plan inquiry chat problems for Medical Billing Services?

Incomplete or outdated policy documents are the top cause – if your knowledge base lacks specific plan terms, eligibility rules, or late-fee details, the bot will give vague answers or say it doesn’t know. Another common issue is relying on the bot to answer account-specific questions without a custom action to collect and route those inquiries to a human. Finally, not testing with real patient language means the bot may miss trigger phrases for the payment plan action, leaving patients confused.

How do I improve payment plan inquiry chat for Medical Billing Services?

Keep your uploaded documents current: update them whenever your billing policies or payment terms change. Add a custom action that captures the patient’s intent and preferred plan in the chat, then hands off the data to a billing team member through a webhook or inbox notification. Use the Insights tab to see which payment questions come up most and fill in any missing content. And periodically review the bot’s answers in the conversation inbox – if you notice a pattern of “I don’t know,” add that specific detail to the knowledge base.