Compare chatbot costs per useful outcome

The table lists six cost components of a chatbot quotation with their typical contents and a common unresolved question for each.

Add tax, currency, minimum term and billing period. A monthly equivalent tied to an annual commitment differs from a cancellable monthly agreement. Swiss quotations should not mix CHF and EUR amounts without a defined conversion assumption.

Freshdesk. One published example illustrates the issue: Freshdesk lists Freddy AI Agent packs of 100 sessions for US$49 and distinguishes session durations for chat and email. This is an add-on component, not the total price of an operational chatbot. Official add-on documentation, checked on 26 September 2026; local invoicing, tax and the actual contract require separate confirmation.

Zendesk. Zendesk documents resolution tiers introduced in May 2026, while some accounts may still use its earlier billing platform. Your account's definition matters more than a single figure copied from an old comparison.

Same example dataset. Ask each supplier to price the same example dataset: a normal answer, clarification, staff transfer, repeat contact and test usage. Ten messages may constitute one session without representing a successfully completed task.

Baseline. A fictional business handles 2,000 eligible requests monthly. Each currently takes five minutes. At an internal fully loaded rate of €36 per hour, the handling cost is €6,000.

With the bot. In the planning example, the bot adequately completes 1,100 requests. The remaining 900 each require four minutes of staff time, costing €2,160. Add €500 for platform and usage, €300 for knowledge maintenance and quality review, and €200 for integration operation. Total ongoing cost is €3,160.

Difference. The calculated difference is €2,840 per month. An assumed €8,000 setup would be covered after approximately 2.8 months if all assumptions remained constant. This is not a forecast: volume, task suitability, quality, use of staff capacity and ongoing expense require operational evidence.

Benefit. Describe the benefit accurately too. Released staff capacity does not automatically become immediate cash savings. It may enable other work or reduce waiting times. Whether payroll cost actually decreases is a separate business decision.

Fewer completions. If adequate bot completions fall from 1,100 to 700, 1,300 requests remain. At four minutes and €36 per hour, residual work costs €3,120. With the same €1,000 of other expenses, operation costs €4,120 and the difference from the baseline falls to €1,880.

Longer residual handling. If average residual handling also rises to six minutes, staff cost becomes €4,680 and total cost €5,680. The monthly advantage is then only €320. The nature and quality of the remaining cases can matter more than a small discount on model calls.

The table sets the three synthetic scenarios side by side.

Peak and low-volume months. Also model a peak month and a low-volume month. Fixed fees remain when demand falls. Higher use may trigger overages, more seats or a different plan. Determine whether reaching a limit stops the capability or automatically creates additional charges.

Numerator and denominator. Divide all relevant costs by a clearly defined number of completed requests. State whether the denominator includes only bot completions or all service completions. Putting platform fees in the numerator and every customer contact in the denominator can make the service appear inexpensive without measuring its actual contribution.

Repeat contacts. Include repeat contact about the same issue. If a customer phones after a supposed resolution, that handling effort belongs to the original task. Review some automatically completed cases for correctness; a high reported resolution rate is not a substitute for quality evidence.

Comparable case mix. Keep the baseline and automated case mix comparable. A month of basic delivery questions should not be compared directly with a later period dominated by disputes and complex complaints.

What a trial shows. Free access can demonstrate the interface and simple answers. It does not prove that personal data access, required channels, permissions, staff handover or exports are included in the eventual plan. Use appropriate test data and inspect the terms before processing real customer information.

Exit costs. A meaningful comparison includes acceptance work, remaining exceptions and the later exit route. Ask whether sources, conversations and configuration can be exported, and who removes the integration at contract end. Transition costs belong in a multi-year purchasing decision.

Test consumption. Confirm whether internal previews and automated evaluation runs consume the same paid allowance as customer use. Budgeting only visible customer conversations can overlook testing and maintenance consumption.

Collect representative requests, current handling times, intended channels and required system actions. Request separate setup and ongoing quotations, with an explicit counting rule for each variable item. Name the people responsible for maintenance and residual service.

The final decision should follow a bounded, transparently evaluated test of those tasks. What matters is whether the chatbot produces suitable outcomes at an acceptable total cost. A low starting price alone cannot establish that result.

These pages cover the service overview and the related decisions about chatbot costs:

• AI chatbots for customer service
• Compare chatbot providers for customer service
• Calculate WhatsApp chatbot platform and operating costs
• Measure resolved chatbot requests and remaining work

For a workflow review, start with the task, systems involved and typical exceptions. Discuss it with modernagents.