Free tool · runs in your browser

What a support bot really costs per resolved ticket

The pitch compares the bot's price per conversation with an agent's wage. It leaves out the tickets the bot hands over, the customers who come back after a "resolved" chat, and the credits you give to calm them down. Put in your numbers and see the cost of a ticket that is actually resolved, with and without the bot.

Nothing you type leaves this page.

Start from an example

Illustrations to show how the result moves, not industry benchmarks. Use your own figures where you have them.

Your support desk

Tickets a monthEvery new conversation, all channels the bot coversAgent cost ($/hour)Wage plus benefits, tools and management, per hourMinutes per ticketAverage time an agent spends on one ticketCome back after an agent (%)Tickets a person closed that the customer reopens

The bot

Resolved by the bot (%)The rate on the vendor's dashboardHanded to a person (%)Asked for a human, or the bot gave upCome back after the bot (%)"Resolved" chats where the customer contacts you again within a weekExtra minutes after the botReading the transcript, an annoyed customer explaining againBot price ($ per resolution)From the vendor's price list or your own model billBot fixed cost ($/month)Platform fee, plus the hours someone spends keeping its answers rightCost of a failed answer ($)Average credit or refund when a "resolved" chat comes back; 0 if none

Charged per resolution Charged per conversation

The gap this tool is about

In 2024 Klarna said its assistant did the work of 700 agents. A year later it was hiring people again, and its CEO said cost had been "a too predominant evaluation factor". Commonwealth Bank of Australia cut 45 support roles on a forecast of what its voice bot would do, then reversed the cuts when call volumes rose instead. Neither switched the bot off. Both had counted what it answered and not what it resolved. The full story, with sources, is in Klarna, Duolingo and the AI-first memo.

Pricing per resolution sounds like it fixes this, and it helps. Read the definition, though. Intercom, for one, counts an outcome for its Fin agent when the customer confirms the answer helped or when "they don't ask for more help after Fin responds". A customer who gives up and emails you the next morning can be a paid resolution and a new ticket. That is what the come-back rate above is for: measure it for a month before you believe any other number.

If your bot talks to customers in the EU, it also has to tell them it's a bot since August 2026. We wrote up what the EU AI Act asks of a small software vendor.

Source for the outcome definition: Intercom pricing page (read October 2026). A model, not a forecast: it leaves out training time, peak-hour staffing and customers who quietly leave. More free tools: biz59.net/tools.