Customer Support
AI ROI Calculator: Work Out the Payback

01. The payback formula

Support AI ROI is simple once you separate the two effects. The time saved is the share of contacts the AI fully resolves, multiplied by the average handling time, multiplied by your loaded cost per hour. The cost is the subscription plus any overage. Payback in months is the subscription divided by the monthly saving. The formula only holds if the AI genuinely removes work rather than moving it, so a metric most vendors do not report, deflected contacts, matters more than any feature list.

02. The four inputs

You need four numbers, all of which you can get from your own support tool in under ten minutes. If you cannot produce them, you cannot calculate payback, and you should not be signing anything yet.

• Contacts per month, split by channel.
• Average handling time per contact, including wrap-up.
• Loaded hourly cost of an agent, including benefits and overhead, not base salary.
• Percentage of contacts that are routine and fully resolvable without a human.

03. Worked example

A three-person service business takes 600 contacts a month, mostly phone and email, at an average handling time of nine minutes. Loaded cost is $34 an hour, which is $28,000 a year per seat before benefits. If an AI receptionist resolves the 55 percent of calls that are routine, that is 330 contacts a month at nine minutes, or 49.5 hours saved per month, worth $1,683. Against a $79 flat plan, payback is under two weeks. Against a $0.99 per resolution model that only touches chat, the arithmetic does not work at all, because the savings are in the phone channel and per-resolution pricing only counts AI answers.

04. When it does not pay back

Two cases. First, if under roughly 20 percent of your contacts are routine, the deflection ceiling is too low to matter and no AI product will show a return. Second, if your highest-volume channel is email with long, case-specific threads, AI drafts will be edited heavily enough that the saving is drafting time rather than handling time, which is a much smaller number. Measure whether agents are accepting drafts unchanged; if acceptance is under about 40 percent, the knowledge base is the problem, not the model, and no purchase fixes it.

• Under 20 percent routine contacts: no AI product will show a payback.
• Long case-specific email threads: expect draft acceptance under 40 percent.
• Low draft acceptance means fix the knowledge base before buying anything.

05. Compare that against per-resolution pricing

Per-resolution pricing looks cheap per unit and scales with exactly the volume that makes the AI valuable. At 1,000 automated resolutions a month, Intercom Fin at $0.99 is $990, and Zendesk AI at $1.50 is $1,500, both before seats. A flat team plan that covers the same volume is $79. The comparison that matters is not cost per resolution, it is total monthly cost at your volume, which is why flat pricing wins decisively for anyone whose contact volume is high enough to benefit.