YundaDeskA few weeks before a big sale, you probably do this: pull last year’s ticket volume, multiply it by this year’s expected traffic, and rough out a support cost estimate. If your AI support tool bills per conversation or per resolution, that estimate delivers an uncomfortable message: your bill scales with traffic, right when a traffic spike is the last thing you want your budget to be hostage to.
This is not scare talk. It is a structural feature of the pricing model. Per-usage billing sounds fair on paper — pay for what you use — but for cross-border sellers, “what you use” is exactly the variable you control the least. A flash sale, a livestream spike, an algorithm change on the platform side, and inquiry volume jumps overnight. When your bill tracks the variable you can least control, it is worth doing the math out loud.
How per-conversation and per-resolution billing actually work
Two common versions of usage-based pricing show up in this space.
Per conversation: every new session the AI handles gets billed once, whether it wraps up in three lines or drags on for ten back-and-forths. Simple in theory, except what counts as “one conversation” — does a customer replying ten minutes later start a new session, does switching topics count as a new one — varies by platform, and that definitional fuzziness is itself a cost risk.
Per resolution: you only pay when the AI “resolves” an inquiry. It sounds fairer — no fix, no charge — but who decides what counts as resolved, and how, is usually a call the vendor makes unilaterally, one you can rarely audit line by line after the fact. Worse, this model creates a built-in incentive to label more conversations as resolved, rather than an incentive to actually get the answer right.
Either way, the underlying pattern is the same: your spend climbs linearly — sometimes faster — with inquiry volume, and inquiry volume is not a number you can lock in ahead of time.
The Hidden Bill Behind Per-Conversation AI Pricing: start the cost discussion with a productivity baseline
Why the bill spikes exactly when peak season hits
Day-to-day inquiry volume is a relatively stable baseline. The trap in usage-based billing tends to show up specifically during peak season.
There is a mismatch here that is easy to miss: a spike in inquiries does not mean your margin is spiking proportionally. Margins during a sale are already thinner because of the discounts, and now support costs climb right along with inquiry volume — an unplanned tax landing exactly in the window you most need cash flow to stay predictable. Worse, you cannot lock this number in ahead of time — you only find out how much you actually spent once the invoice arrives, which makes annual budgeting and margin planning that much harder to pin down.
The hidden cost is not just money — it is decision quality
Usage-based billing does more than move the number on your invoice. It also quietly distorts decisions.
- You start second-guessing whether to build out the knowledge base further to cut AI-to-human handoffs — but now the reason is “reduce the bill,” not “make the customer experience better,” and those two goals do not always point the same direction.
- You hesitate before peak season on whether to temporarily turn off AI coverage on certain channels, worried about blowing the budget — right when that coverage is most needed to fill the staffing gap.
- Your finance team ends up treating the support software line as a variable cost rather than a fixed subscription line like most other SaaS tools, which is not great for a small team’s cash-flow planning.
None of this shows up directly on the invoice, but it quietly warps how your team makes decisions — decisions that should be centered on customer experience end up factoring in billing mechanics instead.
A flat plan with included AI credit: pricing the variable out
YundaDesk takes a different approach: plans include AI credit, with no per-conversation or per-resolution surcharge. You pay a predictable subscription price. AI support handling customer inquiries and Yuna handling merchant-side data questions and setup both draw from the AI credit included in your plan — not a separate line item billed after each conversation.
That structure delivers three concrete benefits:
| Dimension | Per-conversation / per-resolution | Flat plan with included AI credit |
|---|---|---|
| Bill predictability | Swings with inquiry volume, hard to forecast for peak season | Fixed subscription price, known in advance |
| Peak-season decisions | Have to weigh “coverage” against “cost” | No hesitation about turning on channels |
| Financial planning | Support cost is a variable line | Support cost is a fixed line |
This is not to say usage-based pricing is inherently bad or flat plans are inherently good — different business models suit different pricing structures. But for cross-border sellers whose inquiry volume swings hard and clusters around a handful of peak weeks, a bill that does not double the moment traffic doubles is itself a form of certainty — and certainty tends to be undervalued right up until the moment you need it most.
Three questions to ask beyond the price sheet
If you are evaluating vendors, ask three questions before signing that go beyond the headline price:
- If inquiry volume doubles during a sale, roughly what would my bill look like that month? Can you give a concrete estimated range?
- Who decides what counts as a “conversation” or a “resolution” for billing purposes? Can I see a line-item breakdown to audit it myself?
- If I want to temporarily turn off AI coverage on one channel to control cost, does that affect customer experience or data continuity on other channels?
Working through these three questions will usually make it clear whether a vendor is sharing the uncertainty of peak season with you, or passing all of it on to you.
What to weigh beyond pricing
Pricing is one dimension of vendor selection, not the whole picture. Just as important: does a human back up the AI when it gets something wrong, does anything the AI learns require your confirmation before it goes live, and does anything involving money — like a refund — require human approval. Those governance details matter as much as the pricing model for whether you sleep well during peak season. If you are still comparing platforms systematically, this piece walks through a fuller evaluation framework: how to choose an AI support platform for cross-border selling.
Worth bookmarking alongside it: a full walkthrough of how to pace peak-season support itself, in the peak season support playbook.
Bill predictability does not feel like much on a quiet month, but it becomes very visible the moment a big sale wraps and the invoice lands. Better to run the math now than to be surprised later — peak season already gives you plenty to worry about, and your support software bill should not be one of them. You can see how YundaDesk’s plans work on the pricing page.