Two smiling female call center agents with headsets engaging in teamwork.
Photo by Yan Krukau on Pexels

Routing & Automation

Part of Support software costs

Estimating AI feature charges from actual case volume

Turn case counts into an AI cost forecast using the provider's session, outcome or resolution rules.

Start with the cases the proposed AI feature could encounter, then estimate how many meet the provider's billable-event rule. Total ticket volume alone is a poor multiplier: some cases never reach AI, and sessions, outcomes and resolutions are counted differently.

Build a case-led forecast

Choose a representative period and group cases by channel and issue type. Record which groups would be exposed to AI, how many customers are likely to interact with it and which interactions may need a person. Keep a busy period separate. Use observed case counts for the starting volume; label any AI-use or billable-share estimate as an assumption.

Forecast row / Count or term to record

Eligible cases
Cases within the proposed AI rollout
AI interactions
Cases in which the customer would encounter and use AI
Billable events
Interactions meeting the provider's rule
Allowance or minimum
Included units, committed volume or pack size
Charge
Quoted rate, currency and billing period

A handoff is not automatically free or billable. Check the provider's usage report definition and whether the underlying conversation can be inspected.

Case-led forecast rows to record

  • Eligible casesCases within the proposed AI rollout
  • AI interactionsCases in which the customer would encounter and use AI
  • Billable eventsInteractions meeting the provider's rule
  • Allowance or minimumIncluded units, committed volume or pack size
  • ChargeQuoted rate, currency and billing period

Apply the actual event rule

Freshdesk Freddy AI Agent bills sessions. An email AI session spans 72 hours from the first customer email; a chat AI session in Freshdesk Omni spans 24 hours from conversation start. Eligible new customers get a one-time allocation of 500 sessions.

Additional session packs are added to your account for the current billing cycle when purchased. Forecast sessions for the intended product and channel, not one session per ticket by default.

Help Scout AI Answers charges current paid-plan customers for a qualifying resolution: AI Answers gives a full answer and the customer does not ask for more help after its last answer. It charges once per qualifying session, even if AI answers several questions. A greeting or clarification alone is excluded. New customers have a trial period, so distinguish introductory treatment from ongoing charges.

Intercom Fin is priced per Fin outcome. Fin used with another helpdesk also has a minimum commitment. Identify the quoted arrangement and event types before estimating a bill.

Zendesk AI agents use automated resolutions. Its documented newer tier model distinguishes assisted escalation, contained resolution and verified resolution: the first two do not draw from the resolution allowance, while verified resolution does. An account may be on a different resolution platform, so obtain the applicable rule and allowance before using those tiers in a forecast.

Provider billable-event rules to compare

  • Freshdesk Freddy AI AgentBills sessions. An email AI session spans 72 hours from the first customer email; a chat AI session in Freshdesk Omni spans 24 hours from conversation start. Eligible new customers get a one-time allocation of 500 sessions.
  • Help Scout AI AnswersCharges current paid-plan customers for a qualifying resolution: AI Answers gives a full answer and the customer does not ask for more help after its last answer. It charges once per qualifying session, even if AI answers several questions. A greeting or clarification alone is excluded. New customers have a trial period.
  • Intercom FinPriced per Fin outcome. Fin used with another helpdesk also has a minimum commitment.
  • Zendesk AI agentsUse automated resolutions. Its documented newer tier model distinguishes assisted escalation, contained resolution and verified resolution: the first two do not draw from the resolution allowance, while verified resolution does. An account may be on a different resolution platform.

Calculate and reconcile a range

For each issue group, estimate low, central and high billable-event shares of the AI interactions. Apply the quote's included units, pack sizes, minimums and rates. Keep any agent-facing AI licence as a separate seat charge. The range is a planning estimate, not a measured resolution rate.

After a limited rollout, compare the usage report with a sample of cases. Check repeated contacts, handoffs and bot-only conversations against the documented charging rule, then update the forecast from observed billable units before widening exposure.

Calculate and reconcile an AI charge range

  1. Estimate low, central and high billable-event sharesFor each issue group, estimate shares of the AI interactions.
  2. Apply included units, pack sizes, minimums and ratesUse the quote's commercial terms for the intended product and channel.
  3. Keep agent-facing AI licences separateTreat agent-facing AI licences as a separate seat charge.
  4. Run a limited rolloutCompare the usage report with a sample of cases.
  5. Reconcile and updateCheck repeated contacts, handoffs and bot-only conversations against the documented charging rule, then update the forecast from observed billable units before widening exposure.

More in Routing & Automation