AI Customer Support Pricing: Per Conversation, License or Project

AI customer support pricing usually comes in three flavors: per conversation, a monthly license or a custom project. None is always the cheapest. What decides it is how many conversations you handle each month and how much the assistant must be adapted to your business. Below, what each model includes and a simulation to compare them.

The three pricing models

Per conversation (or per resolution)

You pay for what you use. There is no big fixed fee, and cost rises and falls with demand. It suits a start, low volume or seasonal peaks. The risk is a bill that grows without warning once support takes off.

Monthly license

A fixed price per plan, normally with a conversation limit included and overage fees beyond it. It gives budget predictability. The risk is paying for capacity you do not use, or blowing past the limit and paying extra.

Custom project

A setup fee plus a smaller monthly maintenance fee. The assistant is designed around your processes, integrated with your systems and trained on your data. It costs more upfront, but cost per conversation falls sharply at high volume, and you have more control.

What moves the price most

• Monthly conversation volume and average conversation length.
• Number of channels, such as website, messaging apps and email.
• Integrations with ERP, CRM or scheduling to look up and record data.
• Quality and amount of source content, like catalogs and manuals.
• Handoff to a human with full history.
• Security level and privacy requirements.

Illustrative simulation by volume

The numbers below are hypothetical, only to show the logic. Assume per-conversation pricing of $0.25 to $0.75, a license between $150 and $500 a month with an included limit, and a custom project with maintenance of $300 to $800 a month, plus setup.

• 500 conversations a month: per conversation lands between $125 and $375; usually the cheapest.
• 2,000 conversations a month: per conversation reaches $500 to $1,500; the license starts to compete.
• 10,000 conversations a month: per conversation can hit $2,500 or more; a license with overage or a custom project tends to win.

The logic is that usage pricing wins early, licenses win in the middle and projects win at high volume or when integration is heavy. Replace these with the real quotes you receive.

How to budget without surprises

1. Pull your real volume for the last three months by channel.
2. Find out how many conversations are simple questions and how many need system lookups.
3. Ask each vendor for the total cost at your current volume and at double it.
4. Ask what happens past the limit and how the bill is calculated.
5. Include your team’s hours for reviewing and updating the knowledge base.
6. Compare with your current cost per human-handled contact as a reference point.

For a simulation with your own volume, see how we work in AI and automation and ask for a proposal with ranges per scenario.

Frequently asked questions

Which pricing model is best for beginners?

Usually pay-per-use or a basic license, since both need little upfront investment and let you learn your real volume.

Are there hidden costs in AI customer support?

Yes: integrations, extra channels, conversation overage and the staff hours to keep the knowledge base current. Ask for everything itemized.

When does a custom project pay off?

When volume is high, there are many integrations or specific rules, or support is central to your operation.