Train AI to Understand Your Pricing and Plans

Why Pricing Data Is Critical

"How much does it cost?" is among the top three questions customers ask any business. If your chatbot cannot answer pricing questions, it fails at one of its most important jobs. Worse, if it answers with outdated or incorrect pricing, you create real problems: customers expecting a price you no longer offer, or prospects leaving because they think you are more expensive than you actually are.

Getting pricing right in your training data requires two things: completeness (cover all plans, tiers, and options) and maintenance (update immediately when prices change).

How to Write Pricing Training Data

Describe Each Plan or Tier Completely

For each pricing option, write a paragraph that includes the plan name, the price, the billing cycle, what is included, what is not included, and who it is best for. Customers need all of this information to make a decision, and the chatbot should be able to provide it all in one answer.

Example: "The Starter plan costs $19 per month when billed monthly, or $15 per month when billed annually ($180 per year). It includes up to 500 monthly transactions, 2 team member seats, email support, and access to the basic reporting dashboard. It does not include the advanced analytics module, API access, or phone support. The Starter plan is best for small businesses processing fewer than 500 orders per month."

Include Explicit Comparisons

Customers rarely ask about just one plan. They want to know the difference between plans so they can choose the right one. Write comparison paragraphs that clearly state what each upgrade adds: "Upgrading from Starter ($19/mo) to Professional ($49/mo) adds unlimited transactions, 10 team member seats, API access, advanced analytics, and priority email support. The biggest value of the Professional plan is removing the 500-transaction limit, which matters if your business processes more than 500 orders per month."

Cover Common Pricing Questions

Include answers to these questions in your training data:

• What does each plan cost (monthly and annually if applicable)?
• What is included in each plan?
• What is the difference between plans?
• Is there a free trial or free tier?
• Can I upgrade or downgrade at any time?
• Are there setup fees or hidden costs?
• Do you offer discounts for annual billing, nonprofits, or education?
• What happens if I exceed my plan limits?
• How does billing work (when am I charged, what payment methods)?
• Can I cancel at any time?

Handling Complex Pricing

Usage-Based Pricing

If your pricing depends on usage (per message, per transaction, per API call), include specific examples: "AI chatbot messages cost 2 to 15 credits per response depending on the AI model selected. A typical support chatbot using GPT-4.1-mini costs about 3 credits per message. At 1,000 credits = $1, a business handling 100 chatbot conversations per day would spend approximately $9 per month on AI model costs."

Tiered Pricing

For tiered pricing, describe each tier with its thresholds: "The first 1,000 emails per month are included in the base plan at no extra cost. From 1,001 to 10,000 emails, the cost is $0.001 per email ($1 per 1,000). Above 10,000 emails, the cost drops to $0.0005 per email ($0.50 per 1,000)."

Custom or Enterprise Pricing

If you have custom pricing for large customers, be honest about it: "Enterprise pricing is custom based on your volume and requirements. Contact our sales team for a quote." Do not leave this information out, as customers who need enterprise pricing will ask about it.

Keeping Pricing Data Current

Pricing is the most time-sensitive training data in your knowledge base. When you change prices, update the training data on the same day. Old pricing in the chatbot is not just unhelpful, it can create legal and customer service problems if customers rely on quoted prices.

Tag all pricing-related embeddings with a "pricing" tag so you can find and update them quickly. Set a calendar reminder to verify pricing data monthly, even if you do not think anything has changed. See How to Keep Your AI Training Data Up to Date.