At real volume, it can become a six-figure annual line item.
AI pricing often looks simple in a proposal: a few cents per minute, a price per conversation, or a small charge for each successful outcome. The surprise usually comes when that rate meets actual production volume.
The quoted price may be completely accurate. The quick mental math usually is not.
Example: 10,000 minutes a day × $0.03
$300 a day • about $9,000 a month • more than $100,000 a year
Compare vendors using your workload.
One vendor may charge per seat, another per token, minute, conversation, resolution, or outcome. Those quotes are not directly comparable. Apply each pricing model to your real volume over 12 and 36 months.
Model the peak—not only the average.
Usage-based pricing moves part of the risk to the buyer. A seasonal spike, a busy sales day, or stronger employee adoption can increase the bill even when the system works exactly as promised. Ask every vendor to price your busiest realistic month.
Count the bill behind the bill.
The usage charge may not include storage, integrations, retries, implementation, support, quality reviews, change management, or the people still needed for contacts the AI cannot resolve. Those costs belong in the comparison too.
Watch for the success tax.
If you pay per resolution or outcome, improving automation can increase the number of billable events even when customer volume stays flat. Ask the vendor to show how the invoice changes as performance improves.
Three questions worth asking
- What exactly is the billable unit—and who counts it?
- What will our busiest realistic month cost?
- What is not included in the quoted price?
One useful step this week:Â Take the last 90 days of actual volume and ask each vendor to model the next 12 and 36 months, including the peak month and every excluded cost.
Bonus tip:Â Ask for a sample invoice at your average volume and another at your peak. A side-by-side invoice often exposes fees that a pricing slide does not.
Trying to make sense of an AI quote?
Ian would be glad to compare notes on the assumptions that can turn a promising pilot into an uncomfortable production bill.
Talk with Ian: Schedule a Conversation with Ian
