How much does AI-assisted testing cost, and what are you paying for?
With AI-assisted testing, you are not paying only for access to a language model. The cost also includes scenario design, review of generated code, data preparation, execution infrastructure and, depending on the chosen solution, licences. A useful estimate therefore separates the one-off creation of the suite from its ongoing operation and maintenance.
What is the problem?
Proposals can differ significantly. One solution uses the model only when creating a test, while another uses it during every run, evaluation and repair. Without this distinction, a low initial price can conceal recurring model usage or a platform subscription.
Model costs are usually based on the provider’s price list and the volume of inputs and outputs processed. You also need to include the time of the person who defines expected results, reviews the draft and investigates failures. Open-source tools may reduce recurring licence costs, but operation, updates and support remain part of the total cost.
How we address it
We divide the price into components that can be estimated separately:
- Test creation and review: the number of scenarios, their complexity and the application’s readiness.
- Model usage: the selected model, the amount of context and the number of drafts or repairs.
- Operation: CI/CD, test environments, browsers, devices and retention of outputs.
- Maintenance: how frequently the product changes, the stability of test data and the time required to evaluate failures.
- Licences: only where the selected model, service or platform requires a paid plan.
If we use AI only for creation and assisted maintenance, the output can be standard Playwright code in your repository. This reduces dependence on a particular model because the approved tests run independently.
We explain the components of the underlying automation cost in How much does automated testing cost?. AI may shorten some activities, but it does not remove the need for a test strategy or expert review.
What to watch out for
Compare the total cost over the chosen period, not only the cost of creating the first test. Also check whether the model is used during every run, who owns the resulting code and whether the suite can operate without a particular platform.
What you gain
- A budget divided into creation, model usage, operation, maintenance and any licences.
- The option to use open-source components where they provide portability and reduce dependence on a supplier.
- A better basis for deciding whether the time saved during creation outweighs the new operating costs.
Next step
Choose one representative scenario and measure both model usage and the time required for review. A no-obligation consultation is available for an estimate based on your scope.
The Pricing page explains how we calculate prices for our other services.