Showing AI Costs Before Customers Spend Credits

AI features are wonderfully flexible, and that flexibility makes their cost hard to predict. A short paragraph uses very little. A complete website draft with several pages can use a great deal. For an organization with a careful budget, an unexpected drop in their credit balance feels like a broken promise, even if every individual step was correctly priced.

So we design our AI features around a simple principle: show the cost before the work, and confirm before spending.

An estimate first

Before a larger AI task begins, such as generating a website draft or filling in search settings, the feature shows an estimate of how many credits it is expected to use. The estimate is based on the size of the task and the typical usage recorded for that feature.

An estimate is not a guess presented as a fact. It is labeled as an estimate, and the interface explains that the final amount depends on the actual work.

Confirmation before spending

After seeing the estimate, the person confirms they want to proceed. Nothing is spent until they do. If they decide the task is too expensive, they can adjust it, shorten it or come back later.

For tasks that might need more credits than the account has, the interface says so before starting and offers a way to add credits, rather than starting work it cannot finish.

Measured usage after

When the work runs, we measure the actual usage and charge that amount, recording a clear entry in the account's credit history. The record links the charge to the specific task, so the organization can see exactly what each credit was spent on.

Fresh approval for new work

Approval applies to the task that was estimated. If a task fails partway and someone wants to try again, or wants a larger version, the new run needs a new estimate and a new confirmation. An earlier "yes" does not authorize later spending. This rule protects organizations from automated retries or accidental repeats quietly using credits.

Work that was already completed and paid for is reused when possible rather than generated again, so a retry of a partly finished task does not charge twice for the finished parts.

Uncertain attempts are not charged twice

If an AI request ends without a clear result, we do not automatically repeat it. The attempt is recorded as uncertain, and it is not charged again or replayed until the outcome is known. You can read about that approach in locking uncertain outcomes.

Every AI feature declares its billing

None of this works unless the platform knows how each AI feature is billed. That is why every AI use is listed in a registry with its billing rules before it can run, as described in every AI call is registered before it runs. Customer features charge measured credits. Internal work and support help during onboarding are not charged to customers.

Credits and money stay connected

AI credits are bought with real money through the account's wallet, and the price of each credit package is shown before it is purchased. That way, an estimate in credits can always be understood in money terms, and an organization can decide how much to set aside for AI work just as it would for any other expense.

Estimates improve with real data

Every completed task records its actual usage. Over time, those measurements make estimates more accurate, because they reflect how a feature really behaves rather than how we predicted it would. When an estimate turns out to be consistently high or low for a feature, we adjust how that feature estimates, and the next customer sees a more realistic number.

A rule we apply to ourselves

Our own team follows the same principle when using paid AI features for internal work: the cost is known before the work runs, and new paid work needs a fresh decision. Treating our own spending carefully keeps us honest about what we ask of customers, and it keeps costs visible in the same way as every other expense. You can read more about measured usage in metered usage with a prepaid wallet.

Transparency builds trust

Organizations that trust the pricing use AI features more thoughtfully and more often. Organizations that feel surprised stop using them, or worse, stop trusting the platform. Clear estimates, explicit confirmation and honest records are how we earn that trust for features in VoVo.

Frequently asked questions

How do I know what an AI task will cost before I start?

Larger AI tasks show an estimate of the credits they are expected to use, and nothing is spent until you confirm.

Will I be charged again if an AI task fails?

A failed task is not charged for work that did not complete, and starting it again requires a new estimate and confirmation.

Where can I see what my AI credits were used for?

Each charge is recorded in your account's credit history and linked to the task that used it.