Answer

What are the monthly running costs of an AI automation?

Running cost has three parts: consumption-based model usage, channel costs such as telephony minutes or SMS, and monitoring. It scales with volume, not with headcount, and you set a cap. For most SMEs it is a small fraction of the salary cost of the work the automation replaced.

Last updated: 22 September 2026

What you pay per month and why

Model usage is billed by the amount of text processed, so a month with more emails costs more than a quiet one. Telephony and SMS are per minute and per message at your provider's rate. Monitoring covers alerting, log review and fixing anything that breaks when a connected system changes.

How to keep it predictable

Three controls keep the bill flat: a monthly ceiling that pauses non-urgent processing, routing simple cases to smaller and cheaper models, and caching answers that never change, such as opening hours or warranty terms.

In practice the biggest surprise is never the model, it is a channel: an outbound campaign that sends far more messages than planned. We put a daily send limit on anything that goes out.

Asked next

Does the cost grow if we double our volume?

The consumption part roughly doubles; the build and the monitoring do not. That is why cost per handled item falls as you grow.

Can we host it ourselves to avoid usage fees?

You can, and for high, steady volume it sometimes wins. Below that, your own servers and maintenance cost more than the usage you avoid.