Flat subscription
Enter the monthly plan fee and leave the outcome fee at zero. Check whether plan allowances or overages change at peak volume.
Compare a human-only support baseline with an AI-assisted model. Include the costs that simple ROI estimates skip: unresolved conversations, supervision, and rework.
Vendor-neutral model | Example inputs, not a vendor quote | Updated 2026-08-04 | All amounts are editable and shown in USD
Your inputs
Use a normal month first, then rerun the model for your busiest month.
Exclude queues the agent should not attempt, such as sensitive or high-risk requests.
Use pilot evidence, not a vendor-wide resolution claim.
Include loaded labor and the time spent switching, researching, and following up.
Enter the quoted subscription, helpdesk, or platform charge.
Leave at zero for flat plans. Use the vendor contract definition for outcome-priced plans.
Include QA review, source maintenance, workflow repair, and escalation review.
Use loaded hourly cost, not only the person's base wage.
Count outcomes that appeared complete but later needed human correction.
This is a planning estimate, not a quote or financial guarantee. The starting values are illustrative, not Zeiko pricing or a Fin quote. Validate the inputs against your support records, pilot results, and vendor contract. A completed outcome is not automatically a correct or durable resolution.
Enter the monthly plan fee and leave the outcome fee at zero. Check whether plan allowances or overages change at peak volume.
Enter the contractual fee for each billable outcome. Confirm what counts as an outcome, when it is reversed, and whether minimums apply.
Enter both the base platform fee and per-outcome fee. Add any required helpdesk seats to the monthly platform total.
Use average volume and the completion rate observed in a supervised pilot.
Increase volume, lower the completion rate, and include overtime or seasonal support cost.
Raise rework and supervision assumptions to see how quickly expected savings disappear.
Include the platform fee, any fee per completed outcome, the cost of conversations that still need a person, ongoing supervision time, and rework for incorrect or incomplete AI outcomes.
It is an AI-eligible conversation that finishes without human handling. Use the exact definition in your vendor contract and your own pilot evidence because resolution and outcome definitions vary.
Divide loaded support labor by the number of conversations handled, then account for research, context switching, follow-up, management, and tools that are not already included elsewhere in the model.
No. Cost is one decision input. Test answer quality, handoff, rework, customer impact, action safety, and peak-month behavior before committing or expanding permissions.
Use the evaluation scorecard to test grounding, handoff, action safety, data controls, and proof quality before treating the calculator's completion rate as credible.