How To Price an AI Agent
Price an agent product by calculating cost per active user first, then converting that cost into a break-even floor and a target-margin price.
Question
How should I price an AI agent to hit target margin?
Quick answer
Formula: break_even_price = cost_per_user_month
Target margin formula: required_price = cost_per_user_month / (1 - target_margin_pct / 100)
- Assumption: cost already includes the major workflow terms that execute in production.
- Assumption: pricing and cost are measured on the same monthly unit basis.
- Assumption: willingness to pay is validated separately from cost math.
Example: if cost_per_user_month=$12, break_even_price=$12; at an 80% target margin, required_price=$60.
Fastest Working Method
- Estimate cost per active user/month under realistic usage and retrieval assumptions.
- Use that cost as the 0% gross-margin floor.
- Apply the target margin formula to find the actual required launch price.
- Check whether that price still fits packaging and buyer willingness to pay.
What Moves Required Price Most
- Requests per active user.
- Workflow depth, especially retrieval and fallback paths.
- Model mix across baseline and premium traffic.
- Cache hit rate and repeated-request savings.
Pricing Ladder Example
If cost per user each month is $9, then break-even is $9. At a 70% target gross margin the required price becomes $30, at 80% it becomes $45, and at 85% it becomes $60.
That ladder is useful because it tells you which problem you really have. If buyers will only pay $35, the answer is not "price a little better"; it is usually to lower unit cost, narrow usage, or add limits so the economics fit the price point.
When a Flat Seat Price Breaks
- If p90 usage is several times higher than p50 usage, unlimited seats can hide negative-margin heavy users.
- If premium routing only triggers on a minority of workflows, usage caps or add-ons can protect margin better than one flat price.
- If enterprise accounts create much larger retrieval or support workloads, check account-level economics before copying the SMB seat price.
Pricing Mistakes To Avoid
- Using only vendor token rates and ignoring the rest of the workflow cost stack.
- Setting price from one average user instead of p50 and p90 usage patterns.
- Treating break-even as a healthy launch price instead of the minimum viable floor.
Open companion tool: Break-even Price
Baseline cost inputs: AI Workflow Cost
Related reads: How Much Does an AI Agent Cost?, How to Calculate the Break-even Point for AI Workflows
Run the Calculator
Open the related calculator with your own assumptions before you compare infra, packaging, or rollout choices.
Open Related Calculator