Understanding the true cost per resolution with and without AI in 2025

If you're running a support team, your main goal is pretty straightforward: solve customer problems, and do it well. To see how you're stacking up, many teams lean on a metric called "cost per resolution." It’s a simple way to figure out how much you’re spending to close out each support ticket. Now, AI has stepped in and completely changed the math on this, but not always for the better. Some AI pricing models can look great on paper but leave you with a surprisingly high bill.

This post is all about the true cost per resolution with and without AI. We'll dig into the tricky parts of popular "resolution-based" pricing models (like the one Zendesk uses) and show you a more transparent, predictable way to automate your support.

What is the cost per resolution with and without AI?

The idea of "cost per resolution" has been around for a while. Back when support was a human-only game, the calculation was pretty direct. But with AI joining the team, things have gotten a lot more complicated. The way you pay for your tools can drastically change what this number actually means for your budget.

Calculating the cost per resolution with and without AI: Human agents

For a traditional support team, the formula seems simple: divide your total support costs by the number of tickets your team resolves. But the real question is, what’s included in "total support costs"? It’s a whole lot more than just salaries.

• Direct Costs: This is the obvious one. It covers agent salaries, benefits, and any performance bonuses. In the U.S., a support agent typically earns somewhere between $30 to $40 per hour, but that’s just the starting point.

• Onboarding & Training: Getting a new agent ready to help customers isn't cheap. A solid training program can last anywhere from two to six weeks and set you back $2,000 to $5,000 per agent. All that time, they're on the payroll but not yet closing tickets.

• Infrastructure: Your agents need the right setup to do their job. That means paying for office space (if you’re not fully remote), computers, headsets, and licenses for all your software, like CRMs and help desks.

• Attrition: Agent turnover is a massive hidden cost. When someone leaves, you’re not just paying to recruit and train their replacement; you're also losing the knowledge and speed of an experienced team member.

When you add it all up, the actual cost to have a person resolve a single ticket is often way higher than you'd expect. Most industry folks put the average cost per resolution with and without AI (in this human-only scenario) at about $3.00 to $6.00 per interaction.

Calculating the cost per resolution with and without AI: The AI component

AI is a fantastic way to bring that cost down, but it adds a new wrinkle: the AI pricing model. How you pay for your AI has a huge impact on your final cost per resolution. Today, there are two main ways this works:

1. Resolution-Based Pricing: You pay a fee every time the AI "resolves" an issue without a human needing to step in. It sounds good, right? Paying for results.

2. Usage-Based Pricing: You pay a predictable fee based on a clear metric, like the total number of times the AI interacts with a customer, regardless of what happens in each specific conversation.

The first model is tempting, but it can be a trap. Let's get into why.

Hidden problems with AI resolution-based pricing

"Pay for results" sounds fair on the surface. But in reality, resolution-based pricing is often built on shaky ground, with fuzzy definitions that lead to unpredictable bills and goals that just don't line up. You don't have to look far on public forums to find people who've been burned by this, and their stories all point to a few common headaches.

Problem 1: Flexible 'resolution' definitions inflate costs

Here’s the biggest catch: nobody agrees on what a "resolution" actually is. This gives vendors the freedom to define it in whatever way helps their bottom line, not yours. You can end up paying for chats that weren't helpful at all.

Here are a few common situations that often get billed as a full resolution:

• A customer gets annoyed and just closes the chat window. Is their problem solved, or did they just give up?

• The AI points them to a help article. The user clicks it, but the article doesn't actually fix their problem. That click is often enough to count as a charge.

• The AI completely misunderstands the question, but because the user doesn't ask for a human agent within a certain time (some platforms wait up to 72 hours), it's marked as a success.

This ambiguity means you’re often paying for someone giving up, not for actual satisfaction.

Problem 2: Unpredictable costs and budget impact

When your bill is tied to a metric as vague as a "resolution," trying to forecast your monthly spend becomes a real guessing game. Your ticket volume goes up and down, and customer behavior is all over the place. One busy month could land you a bill that's way higher than you expected.

This model also creates a strange situation: the better your AI does, the more you pay. Let's say you work hard to improve your knowledge base, and your automation rate jumps from 30% to 60%. For the same number of customer conversations, your bill could double. You're essentially being penalized for success, which takes away the incentive to keep getting better.

Problem 3: Misaligned goals

At the end of the day, this pricing model can put you and your AI vendor at odds. Your goal is to have happy customers whose problems are truly solved. But when a vendor's revenue is tied to the number of "resolutions," their goal can become just doing whatever it takes to tick a box and trigger a charge. The focus shifts from the quality of the conversation to hitting a technical metric, which is the opposite of what good customer service is about.

A look at Zendesk’s automated resolutions pricing

Zendesk is a great example of a platform that has gone all-in on the resolution-based model with what they call "Automated Resolutions" (ARs). Let's take a closer look at how it works and why it can be a headache for businesses.

How Zendesk's pricing works

Every Zendesk Suite plan comes with a very small number of ARs per agent each month. Once you burn through that tiny allowance, the overage fees kick in, and they're not small.

• Pay-as-you-go: $2.00 per automated resolution.

• Committed usage: $1.50 per automated resolution.

This means a team that automates an extra 1,000 tickets in a month could be looking at an unexpected bill of $1,500 to $2,000 on top of what they already pay. For teams trying to automate more and more, these costs can spiral out of control fast.

Plan Tier	Included Automated Resolutions (per agent/month)	Pay-as-you-go Cost per Resolution	Committed Cost per Resolution
Suite Team	5	$2.00	$1.50
Suite Professional	10	$2.00	$1.50
Suite Enterprise	15	$2.00	$1.50

What Zendesk counts as a resolution

What makes it even more confusing is the complex logic Zendesk uses to decide what's a billable resolution. According to their own documentation, a resolution can be counted after 72 hours of silence from the user, as long as a language model confirms the AI's response was "relevant."

Think about that for a second. A customer could ask a question, get an answer that doesn't help, and just leave. If they don't reply within three days, Zendesk's system can mark it as a success and charge you for it, even if the customer never said their issue was solved. This is exactly what so many users worry about: paying for conversations that just fizzled out.

This workflow illustrates how a Zendesk AI agent handles a ticket, which can impact the cost per resolution with and without AI.

A better way: Clear pricing with eesel AI

For businesses that are tired of the fuzzy definitions and surprise bills, there’s a much better way to do this. A transparent, usage-based model gives you a clear path to scale your AI automation without worrying about your budget.

Transparent and predictable pricing

We built eesel AI with a different approach in mind. Our plans are based on a simple, predictable metric: the number of monthly AI interactions, where one interaction is just one AI reply or one AI action.

The benefit is pretty obvious: no per-resolution fees. Your bill won't suddenly jump just because your AI is doing a great job. Instead, you get a big pool of interactions for a flat monthly price, which makes it easy to plan your costs. We also offer flexible monthly plans you can cancel anytime, so you’re never locked in.

Plan	Monthly Price (Billed Monthly)	AI Interactions/mo	Key Benefit
Team	$299	Up to 1,000	Predictable cost for startups
Business	$799	Up to 3,000	Scales confidently with your team
Custom	Contact Sales	Unlimited	Enterprise-grade volume and features

eesel AI's public pricing page shows a clear, usage-based model, offering a predictable cost per resolution with and without AI.

Simulate your savings and understand the costs

One of the coolest things about eesel AI is our Simulation Mode. Before you even switch the AI on for your customers, you can run it on thousands of your past tickets. This gives you a super accurate forecast of your potential automation rate, shows you exactly where the gaps are in your knowledge base, and, most importantly, tells you exactly how your usage will fit into our plans.

You’ll know your costs before you commit to anything. With other platforms, you often have to just go live and cross your fingers. eesel AI lets you test everything out completely risk-free.

The eesel AI Simulation Mode helps businesses forecast their cost per resolution with and without AI by testing on past ticket data.

Total control of automation and costs

With eesel AI, you don’t have to go all-in at once. Our workflow engine lets you decide exactly which types of tickets the AI should touch. You can start small, automating simple, repetitive questions, and have everything else go straight to your human agents. As you watch the reports and get more comfortable, you can gradually let the AI handle more. This approach takes the risk out of the whole process and ensures you only automate what you’re ready for.

With eesel AI's workflow engine, you have full control over automation, directly managing your cost per resolution with and without AI.

Stop paying for guesses and understand your true cost per resolution

The true cost per resolution with and without AI is about more than a price tag; it's about predictability, transparency, and being in control. Resolution-based pricing might sound good at first, but it often hides volatile costs behind fuzzy definitions that have little to do with whether your customers are actually happy.

eesel AI's model is the clear, predictable alternative. With flat-rate pricing based on how much you use it and a powerful simulation mode to forecast your return, you can scale your support automation with confidence, knowing your costs are always under your control.

Take control of your cost per resolution

Ready to see how AI can lower your cost per resolution without the billing surprises? With eesel AI, you can connect your helpdesk in just a few minutes and run a free, no-risk simulation on your past tickets to see how much you could save.

Start your free trial today.