Pricing custom AI services is harder than pricing software because there is no standard unit — every client is different, every implementation is different, the savings are highly specific. But there is a framework that works.
First, understand what you are selling. Are you selling implementation (setup, training, integration)? Are you selling ongoing management? Are you selling ROI-based pricing where the client pays a percentage of the time they save? Most successful AI service businesses sell a combination.
Implementation pricing: for a standard customer service AI agent setup with training on their knowledge base, integration with their CRM, and two weeks of active support, $3,000-$8,000 is a reasonable range depending on complexity and your market. A small local business in a Tier 2 city — $3,500. A mid-market business in a major market — $6,000. A complex enterprise integration — more.
Ongoing management pricing: if you are maintaining and optimizing the AI system after launch, a monthly retainer of $300-$2,000 depending on the system complexity and update frequency is standard. This covers monitoring, prompt optimization, retraining, and incident response.
ROI-based pricing: if the client is saving 10 hours per week and you can quantify the value of those hours at their labor rates, you can charge a percentage of the savings. This aligns you with the client's success and handles the objection that your price is high — because the client's savings are higher.
Value-based pricing works when you can prove the value. A legal firm using an AI document automation system that cuts review time in half — that is worth $2,000-$5,000 per month depending on firm size because the leverage is clear.
Do not price by time spent. Price by value delivered.