The Hidden Cost of $0.99 Per Resolution: Why AI Support Gets Expensive Fast
Voxe Team· April 12, 2026The pitch is clean: you only pay when your AI actually solves a customer's problem. No wasted spend on failed attempts. No flat fees for capability you're not using. It sounds like the most honest pricing model in support software.
It isn't.
Per-resolution billing has a structural flaw that only becomes visible once your AI starts performing well. As your AI resolves more tickets — the exact outcome you deployed it to achieve — your costs compound automatically, with no corresponding increase in value. Teams using resolution-based billing have reported watching their support bill jump from $4,000 to over $9,000 a month without adding a single new agent, launching a new product, or seeing any meaningful increase in customer count. The AI just got better at its job. And they got an invoice that reflected it. One user on G2 put it plainly: "You're penalized for your AI performing well." That framing keeps appearing across Reddit threads, Capterra reviews, and G2 listings — in different words, always the same story.
TL;DR
- At $0.99 per resolution, 40,000 AI-resolved tickets per month costs $39,600 — before any human agent costs (based on published per-resolution pricing benchmarks)
- Teams on resolution-based billing routinely report 2–3× bill increases after AI tuning, per community reports across Reddit, G2, and Capterra
- Resolution-based pricing creates a perverse incentive: managers artificially limit AI deployment to control costs, directly undermining the ROI case for the investment
- Entire SEO arbitrage sites exist solely around AI-first support platform pricing pain points — a reliable signal of how widespread and documented this problem is
- Coverage-based models charge for system access, not resolution outcomes — cost stays flat regardless of how well the AI performs
What Does "Per Resolution" Actually Mean?
The definition sounds simple, but it matters more than most teams realize before they sign. According to Gartner's 2024 Customer Service Technology report, 61% of organizations cite total cost of ownership — not initial pricing — as the deciding factor when switching support platforms, usually after experiencing a surprise in their first high-volume month. A resolution, in per-resolution billing, is a ticket closed by the AI without a human agent intervening. The exact threshold varies by platform — some require explicit customer confirmation, others auto-close after inactivity — but the billing logic is identical: AI answers, customer stops responding, the platform counts a resolution, you pay.
That logic sounds fair at face value. The problem is what happens to it over time.
The Resolution Rate Trap
AI support systems get better as you invest in them. You add more knowledge base content, refine confidence thresholds, expand question coverage, and connect more integrations. Each improvement raises your resolution rate. And under a per-resolution model, each improvement raises your bill.
A team that improves AI resolution rates from 30% to 70% on the same monthly ticket volume has effectively more than doubled its resolution billing — with no guarantee that human labor costs decreased proportionally. Headcount is sticky. Bills are not.
This is the trap that makes per-resolution pricing uniquely punishing compared to per-seat models. Per-seat billing penalizes you for adding people. Per-resolution billing penalizes you for making your AI work better. Per-seat pricing at least scales with something you control — headcount. Resolution billing scales with AI performance, which is precisely what you're supposed to be optimizing.
The Compounding Math: How Good AI Performance Becomes a Budget Problem
At $0.99 per resolution — a widely cited benchmark across AI-first support platforms — the math at scale is not subtle. According to a 2024 SaaS Pricing Benchmark report, teams that underestimate consumption-based billing overruns do so by an average of 340% in the first year of a new AI support deployment. Here's what different monthly resolution volumes actually cost:
| Monthly AI Resolutions | Cost @ $0.99/Resolution | Annual Total |
|---|---|---|
| 5,000 | $4,950 | $59,400 |
| 10,000 | $9,900 | $118,800 |
| 20,000 | $19,800 | $237,600 |
| 40,000 | $39,600 | $475,200 |
| 80,000 | $79,200 | $950,400 |
None of those figures include human agent costs, platform subscription fees, or any add-ons. That's purely the resolution billing line. A team handling 40,000 monthly resolutions is paying close to $475,000 per year for the act of AI doing its job — more than the fully-loaded cost of hiring six mid-market customer support agents.
The Spike Problem Is the Real Danger
Average monthly volumes are manageable. Spikes are where the math breaks teams.
A viral product launch, a shipping delay, a single bug that generates thousands of duplicate tickets over 48 hours — these events compound resolution billing fast. A team budgeting for a $5,000 month can find themselves looking at a $15,000 or $20,000 invoice after a single bad week. There's no ceiling. The AI keeps resolving. The meter keeps running.
Teams using coverage-based pricing — including those on Voxe's Team and Business plans — absorb these spikes differently. The plan covers an included annual chat volume. When conversations exceed that quota, additional chats are billed at raw API cost with no markup. A traffic spike is a stress event for your support team's workload. It's not a financial emergency at the end of the month. That distinction matters most exactly when you're already dealing with a crisis.
Why Do Teams Under-Deploy AI to Control Costs?
This is the behavior per-resolution pricing reliably produces — and it's where the model becomes actively harmful, not just expensive. Per a 2024 Forrester report on AI adoption in customer service, 44% of support leaders on consumption-based pricing models said they restricted AI deployment scope to control monthly spend. That's nearly half of teams actively limiting the tool they paid to deploy.
When every AI resolution carries a price tag, support managers start making decisions that have nothing to do with support quality. They set confidence thresholds artificially high so the AI escalates more and resolves less. They avoid enabling AI on high-traffic channels. They restrict the question categories the AI is trained to handle. They run finance reviews before expanding the knowledge base.
The Self-Defeating Loop
All of those decisions are rational under the pricing model. None of them are good for customers or for the business.
The ROI case for AI support is built on automation rate. You deploy AI to handle more tickets with less human labor. Resolution billing means that the higher your automation rate, the higher your bill — which means the lower the net savings. At sufficient scale, the cost curve from resolution billing can approach or exceed the cost of the human labor it displaced.
The 5 metrics that tell you whether your AI support investment is actually working all measure deployment scope and resolution quality — metrics that teams on resolution-based billing are actively managing downward. The model produces the wrong behavior and then measures the damage.
What Does the Community Say About Resolution-Based Pricing?
The signal from review platforms and community forums is unusually consistent — and unusually specific. Reddit, G2, and Capterra all surface similar patterns from teams that have used Intercom's Fin and similar resolution-based AI support platforms. The sentiment is not that the AI performs poorly. It's that the pricing model makes good performance unaffordable.
The most common report: a team runs a controlled pilot, sees promising resolution rates, expands deployment, and receives an invoice that bears no resemblance to what they budgeted. The resolution rate that looked like a success metric in month two becomes the line item that blows the budget in month six. There's no natural plateau — the bill scales with improvement until finance intervenes.
One reliable indicator of how real and widespread this pain is: the existence of SEO arbitrage sites built entirely around AI-first support platform pricing. Sites that rank for "[platform name] pricing" exist because teams are actively searching for alternatives after experiencing bill shock. That's not a niche product complaint. It's a documented, high-intent information gap that independent publishers have built traffic strategies around filling. When a pricing problem is large enough to sustain an industry of competitor-targeting sites, the problem is structural.
The pattern across G2 and Capterra reviews is also specific to timing: teams that initially praised the "fair" outcome-based framing revised their assessments within 6–12 months, after experiencing at least one spike month or significant AI improvement cycle. The model feels fair when volume is controlled. It stops feeling fair when the AI actually works.
What Should a Fair Pricing Model Look Like?
The alternative isn't complicated. The principle is this: charge for access to the system, not for the act of helping customers. Your pricing should align with what you need — the number of chatbots deployed, the annual conversation volume, the number of human helpdesk agents — not with how often the AI closes a ticket successfully.
In a coverage-based model, improving your AI resolution rate doesn't increase your bill. A volume spike during a product launch doesn't generate a surprise invoice. Expanding your knowledge base, adding a new question category, training the AI on billing disputes — none of those decisions carry an incremental cost. The incentive is to deploy AI as broadly as possible, because broader deployment doesn't cost you more.
This is how Voxe's pricing is structured. Plans are built around chatbot count, annual chat volume, and the number of human helpdesk agents — all included in the tier. Human agents don't cost extra per seat. AI resolutions don't cost extra per close. When volume exceeds the included quota, additional chats pass through at raw API cost — no markup, no penalty rate.
| Plan | Monthly Price | Chatbots | Annual Chats Included | Helpdesk Agents |
|---|---|---|---|---|
| Starter | $45 | 2 | 12,000 | 2 |
| Team | $115 | 5 | 60,000 | 3 |
| Business | $245 | 10 | 100,000 | 5 |
| Enterprise | Custom | Unlimited | Unlimited | Unlimited |
Three Questions to Ask Before You Sign a Resolution-Based Contract
If you're evaluating AI support platforms or reviewing at renewal, stress-test the pricing model against actual volume before committing.
- What's my bill if my AI resolution rate doubles? If the answer is a multiplication problem, that's resolution-based billing.
- What happens during a spike month — is there a cap? Get the exact overage rate in writing, not a range or a promise.
- What's the incremental cost of improving AI coverage? Adding knowledge base content, training on new categories, connecting new channels — these improvements should not carry a per-resolution cost. If they do, the model penalizes optimization.
What AI support actually handles well — and how to build the right knowledge base for it is a separate question from pricing. But the pricing model shapes what you're willing to let the AI handle. Get the model wrong and the AI will always be under-deployed, regardless of how capable it is.
FAQ
What is per-resolution pricing in AI customer support?
Per-resolution pricing charges a fee — typically around $0.99 — each time the AI closes a support ticket without human intervention. It's framed as outcome-based billing: you pay for results. At low volume it can feel predictable. At scale, it becomes expensive fast: 40,000 AI-resolved tickets in a month generates a $39,600 bill from the resolution fee alone, before any platform or agent costs. High-volume teams and fast-growing products are most exposed to this model's compounding effect.
Why do per-resolution bills compound over time?
Bills compound because AI resolution rates improve as teams invest in the system — adding knowledge base content, refining thresholds, expanding coverage. A team that improves resolution rates from 30% to 70% on a flat ticket volume has effectively doubled its resolution billing with no guaranteed offset in human labor savings. Every AI optimization that benefits customers also increases the invoice. That structural misalignment compounds across 12-month deployment cycles.
Is Intercom's Fin pricing per-resolution?
Intercom's Fin AI agent has operated on a per-resolution pricing model charging for each ticket the AI closes without human escalation. The pricing has been extensively discussed on Reddit, G2, and Capterra, with teams consistently reporting bill increases after expanding AI deployment or improving resolution rates. Intercom's pricing terms update periodically, so verify the current model directly — but the community pattern around this pricing structure is well-documented and consistent across review platforms.
How do teams typically react when resolution-based bills spike?
The most common response is to artificially constrain the AI. Managers raise confidence thresholds so the AI escalates more (and "resolves" less), restrict which question categories the AI handles, avoid enabling AI on high-traffic pages, and require finance sign-off before expanding coverage. These are rational responses to the pricing structure — and they all reduce customer experience quality. The team ends up paying for an AI capability they're deliberately rationing to protect the budget.
What's the difference between per-resolution and coverage-based pricing?
Per-resolution billing charges for outcomes — each AI-closed ticket costs money. Coverage-based billing charges for access — a flat rate for the system capacity you need. Under coverage-based pricing, a better AI resolution rate doesn't increase your bill. A traffic spike doesn't generate a surprise invoice. You can train the AI on new question categories, deploy it on more channels, and improve the knowledge base without a corresponding cost increase. The pricing incentive is to deploy AI broadly rather than ration it.
How do I estimate my true annual cost under per-resolution billing?
Start with your current monthly ticket volume. Multiply by a conservative resolution rate — say 40% if you're early in deployment. Multiply that by the per-resolution fee. Then model the same scenario at 60% and 70% resolution rates, because those are realistic 12-month targets if you invest in the platform. The gap between your initial estimate and your 12-month projection is the pricing risk you're absorbing. For a concrete alternative, Voxe's 14-day free trial is no credit card required — enough time to test the platform against your actual support volume before any commitment.
Per-resolution pricing isn't dishonest. It's a model that works in limited conditions — low volume, controlled pilots, a ticket base that doesn't grow. It stops working when the AI performs well, when volume spikes, or when the business goal is to automate as much support as possible.
The teams reporting the largest bill increases didn't do anything wrong. They deployed AI broadly, invested in improving it, and watched resolution rates climb — exactly what the investment was supposed to produce. The pricing model just happened to treat that outcome as a cost.