July 23, 2026 · The Evolution team

AI Customer Support Pricing: What Per-Resolution Really Costs You

If you've shopped for an AI support tool recently, you've probably noticed the pricing pitch changed. It's not "per agent seat" anymore, it's "pay only when the AI resolves a ticket." That sounds simple and fair. It's also a lot easier to get wrong than the sales page makes it look.

Most of the major players in ecommerce support, and a wave of newer AI-only tools, have moved to per-resolution pricing over the past year, typically somewhere between $0.50 and $2.00 per resolved conversation depending on the vendor. For a small store, that can genuinely be cheaper than a flat monthly seat fee. It can also quietly cost more than you expected, and the reason almost always comes down to one word: resolution.

The pitch versus the reality

The pitch is straightforward: you only pay when the AI actually solves a customer's problem without a human touching it. No resolution, no charge. That's a fair model in principle, and for a store with seasonal or unpredictable ticket volume, it can be a real improvement over paying for the same number of seats whether you have 200 tickets that month or 2,000.

The catch is that "resolution" isn't a standard, audited term. Each vendor defines it themselves, and the definition is the single biggest lever on your actual bill. Some count a resolution the moment the AI sends a plausible-sounding answer, whether or not the customer's problem was actually solved. Others only count it if the customer doesn't reply again within a set window, which sounds more honest but means an unrelated follow-up message can sometimes reset the count in the vendor's favor, not yours.

Three questions to ask before you sign up

1. Who decides what counts as resolved?

Ask this directly. If the vendor's own AI model marks its own answers as resolved, you're trusting the party being paid per resolution to grade its own work. Look for tools that define resolution around a customer action, no reply within a window, a closed ticket, a positive rating, rather than the AI's own confidence score.

2. What happens to escalations?

If the AI can't answer and hands the ticket to a human (you, or a teammate), do you still pay? Some vendors charge nothing for escalated tickets, which is the fairer version of this model. Others charge a partial fee for the AI's attempt even if it ultimately failed. This single detail can change your effective cost by 20-30% depending on how often your ticket mix needs a human anyway, which for most small stores handling WISMO and presale questions is more often than the demo implies.

3. What's your realistic resolution rate, not the vendor's best-case number?

Vendor marketing pages tend to quote resolution rates from their best customers, often 70-85%. A small store with a narrower, more repetitive ticket mix (shipping status, sizing, return policy) can sometimes beat that. A store with more presale and product-specific questions usually lands lower, closer to 50-60%, which directly changes your monthly bill since you're paying per success, not per attempt.

A worked example

Say you get 900 support tickets a month and are comparing a per-resolution AI tool at $1.00 per resolution against your current $300/month support tool plus your own time.

If the AI resolves 65% of tickets (585 of 900) without a human, that's $585 a month, roughly comparable to a mid-tier flat plan, but now scales up in a busy month and down in a slow one instead of staying fixed. If it resolves 80% instead, the same setup costs $720, which is more expensive in absolute dollars but has also removed 720 tickets a month you or a hired VA no longer have to touch, which is the number that actually matters when you're weighing this against hiring a VA versus automating.

The math only works in your favor if you know your real resolution rate ahead of time, not the number from the sales deck. Ask any vendor for a resolution rate benchmark from stores similar in size and ticket volume to yours, not their platform-wide average.

Where this fits into your bigger stack decision

Per-resolution pricing is one more reason the old model of stacking a separate app for support, a separate one for reviews, and a VA for the rest is getting harder to justify on cost alone, the same conclusion that shows up when you actually total up what your app stack costs per month. A tool priced per resolved outcome only makes sense if you can see, plainly, what it resolved and what it didn't, which is a fair bar to hold any AI tool to, support or otherwise.

What to actually do

Before signing up for anything priced per resolution, pull your last two months of tickets and manually sort them into three buckets: routine and clearly answerable by AI, ambiguous, and clearly needs a human. That single afternoon of sorting tells you more about your real cost than any vendor's calculator, because it's built from your actual ticket mix instead of an industry average.

Want a free audit of your store? Get yours here.