Why Per-Resolution AI Pricing Hurts Support Teams

Per-resolution AI pricing sounds fair on the surface: you pay only when the bot actually resolves a conversation. Many support platforms have adopted it, charging somewhere between roughly one and two dollars every time their AI closes out a ticket. But once you run the model forward, per-resolution AI pricing turns out to create unpredictable bills and a few uncomfortable incentives. This piece makes the case, fairly, for why flat, predictable pricing tends to serve support teams better.

How per-resolution pricing works

Under a per-resolution model, the platform counts each conversation the AI handles to completion and bills you for it. The exact definition of a resolution varies by vendor, but the shape is consistent: more conversations resolved means a bigger invoice. It is usually layered on top of a per-seat or platform fee, so the resolution charge is the variable part of your cost.

The appeal is obvious. It feels like pay-for-value: you only pay when the AI does its job. Vendors like it because revenue scales with usage, and buyers like the idea of not paying for an idle tool. The problem is what happens when usage is not steady, which for customer support is most of the time.

Problem one: the bill is unpredictable

Support volume is spiky. A product launch, a holiday rush, an outage, a viral post, or a shipping delay can multiply your conversation count in a day. Under per-resolution pricing, every one of those resolutions adds to the bill, so the months when you most need the AI to absorb load are exactly the months your invoice spikes. You cannot budget cleanly for a number that moves with events you do not control.

This unpredictability has a real cost beyond the dollars. Finance teams dislike variable line items they cannot forecast. Support leaders end up watching a usage meter instead of focusing on customer outcomes. And the surprise invoice after a busy month sours the relationship even when the AI performed beautifully.

Consider a concrete example. A store that normally handles a few thousand conversations a month launches a popular product and a shipping delay hits at the same time. Volume triples for two weeks. The AI does exactly what you hoped, absorbing the flood so customers still get instant answers. Then the invoice arrives, and the resolution charges have tripled too. The tool worked perfectly and was rewarded with a bill three times larger, in the exact month your margins were already under pressure. That is the per-resolution model in a nutshell: the cost peaks precisely when you can least absorb it.

Problem two: it creates a perverse incentive

Here is the subtler issue. When you pay per resolution, you are financially penalized for success. The better your AI gets at resolving conversations, the more you pay. That is backwards. You want to encourage the AI to handle as much as it can, but the pricing quietly pushes you to wonder whether you should route more conversations to humans to keep the meter down. A pricing model that makes you hesitate to use the product you bought is working against you.

Any time the price goes up precisely because the tool worked well, you have an incentive problem. Per-resolution pricing taxes the outcome you are paying for.

There is also a definitional grey area. What counts as a resolution? If a customer asks one quick follow-up an hour later, is that a new resolution and a new charge? Different vendors answer differently, and the ambiguity tends to favor the vendor. You can end up auditing resolution counts to make sure you are billed fairly, which is time no support team wants to spend.

Problem three: it stacks on top of other variable fees

Per-resolution charges rarely arrive alone. On WhatsApp, you may also be paying Meta's per-conversation fee, sometimes with a markup added by the tool. On SMS, you pay carrier fees. Layer a per-resolution AI charge on top of those, and a single customer interaction can carry three separate variable costs. Each layer is individually defensible, but together they make your true cost per conversation almost impossible to predict in advance. We break down the WhatsApp side in WhatsApp Business API pricing explained.

To be fair to per-resolution pricing

It is not all bad, and it is worth being honest about that. For a business with genuinely low and stable volume, per-resolution pricing can be cheaper than a subscription, because you pay only for the little you use. It also aligns the vendor's incentives with delivering working resolutions rather than shelfware. And for buyers who want to start tiny and grow, the low entry point is attractive. If your volume is small and predictable, the model can suit you.

The trouble is that customer support volume is rarely small and predictable for long. The moment you grow, or hit a spiky month, the model that looked cheap becomes the model that produces the scary invoice.

Why flat pricing serves support teams better

Flat, predictable pricing flips every one of these problems. Your bill is the same whether you have a quiet week or a crisis, so you can budget with confidence and let the AI absorb every spike without watching a meter. The incentive is aligned the right way: you want the AI to resolve as much as it possibly can, because doing more costs you nothing extra. And there is no resolution-counting grey area to audit, because resolutions are not the billable unit.

For a team that depends on customer conversations, predictability is not a minor convenience. It changes behavior. When the price is fixed, you route everything you can to the AI and reserve humans for the conversations that truly need them, which is exactly the outcome good support is supposed to produce.

Flat pricing also makes you a better buyer of your own AI. Under metering, every improvement to the bot, every new intent you let it handle, shows up as more cost, so you second-guess expansion. Under a flat plan, expanding what the AI handles is free, so you are encouraged to push automation as far as quality allows. The pricing model quietly shapes how aggressively you adopt the product, and flat pricing pushes in the direction you actually want.

Questions to ask a vendor about pricing

Before you commit to any AI support tool, get clear answers to these:

• Is the AI metered per resolution or per conversation, or is it included in the subscription? If metered, what exactly counts as a billable unit?
• If a customer sends a follow-up an hour or a day later, does that create a new charge?
• Do you add a markup to WhatsApp, SMS, or carrier fees, or are they passed through at cost?
• What would my bill look like in a month where volume doubles?
• Are there per-seat fees, monthly active contact fees, or other variable charges layered on top?

The answer to that fourth question is the one that matters most. If the vendor cannot give you a number you could budget for, the model is the problem, not the math.

How MessageAgent prices it

MessageAgent uses flat, predictable subscription pricing on purpose. You pick a plan, you know the number, and the AI handles as many conversations as it can across every channel without a per-resolution meter ticking up. The only pass-through costs are the genuine carrier and Meta message fees, and those are passed through at cost with no markup. A busy month does not produce a surprise invoice, and you are never penalized for the AI doing its job well.

That is the wedge against tools that meter you per resolution or mark up your WhatsApp fees: one flat, predictable price, one AI brain across every channel, and a bill you can actually forecast.