Zendesk AI agent metrics to track automated resolutions

So, you’re using AI to handle customer support. It’s supposed to be the key to unlocking 24/7 service and making your team more efficient. But how do you actually know if it's working? More importantly, how do you measure its success without breaking the bank?

For anyone using Zendesk, this all comes down to one thing: automated resolutions.

Getting a handle on the specific Zendesk AI agent metrics to track automated resolutions isn't just about performance reports. Zendesk's pricing is directly tied to this number, so understanding it is crucial for managing your budget. This guide will walk you through what to track, point out some of the tricky parts of Zendesk's model, and show you a way to get more control over your support automation.

What are automated resolutions?

In Zendesk's world, an "automated resolution" is any customer issue that their AI agent solves completely on its own, with no human help needed. This is the main goal of their AI tools and, crucially, the foundation of their billing. When the AI agent answers a question or points a customer to the right help article and the customer is satisfied, Zendesk chalks that up as one automated resolution.

To be fair, they use a Large Language Model (LLM) to double-check that the conversation was actually resolved before it gets counted. This applies to chats, emails, and web forms.

Because you literally pay for each one of these resolutions, knowing exactly what counts is a must for any team on Zendesk. It has a direct impact on your monthly bill and how you set up your AI agents in the first place.

The essential Zendesk AI agent metrics to track automated resolutions

To really understand how your AI agent is doing, you need to look at a few different numbers. Here are the main metrics you’ll find in Zendesk’s analytics dashboards.

Automated resolution rate

This is the headline number. It’s the percentage of all conversations that the AI agent handled from start to finish without needing to pass the ticket to a human.

A high automated resolution rate is a good sign, it means your AI is successfully taking care of common questions. But you can’t look at this number in a vacuum. You have to check it against your customer satisfaction scores to make sure you're actually solving problems, not just closing tickets.

Escalation rate

This is the flip side of the resolution rate. The escalation rate tells you how many conversations the AI agent couldn't figure out and had to pass along to a human agent.

Keeping an eye on this metric helps you see where your AI is hitting its limits. If your escalation rate starts climbing, it could mean you're getting new or more complicated customer questions, there are gaps in your help articles, or your AI agent just needs some fine-tuning.

Deflection rate vs. AI agent-handled conversations

You’ll see a couple of similar-sounding terms in Zendesk's reports, and it’s important to know the difference.

• Deflection Rate: This is a pretty broad metric. A chat is "deflected" if it just doesn't end with an escalation to a human. This could mean the problem was solved, but it could also mean the customer got frustrated, gave up, and just closed the window.

• AI Agent-Handled: This is a much better metric to watch. For a conversation to count as "AI agent-handled," the AI has to understand what the customer wants (a real "intent") and there can't be any attempt to escalate.

If you only look at the deflection rate, you might get a skewed picture of your AI's performance. Focusing on the AI agent-handled rate gives you a much more honest view of where your AI is actually winning.

Bot satisfaction (BSAT)

Numbers are great, but they don't tell the whole story. BSAT is how Zendesk measures customer satisfaction with the AI agent specifically.

After a chat, customers can give a quick rating of their experience. If you have a high resolution rate but a low BSAT score, that’s a huge red flag. It means you’re closing tickets, but leaving a trail of unhappy customers behind. This metric helps you make sure your push for automation doesn’t ruin your customer experience.

Challenges with Zendesk's AI pricing

While Zendesk gives you a few metrics to watch, its pricing model is pretty much all about one of them: the automated resolution. This sounds reasonable at first, but it can create some real headaches for support teams as they grow.

Understanding the per-resolution pricing model

Zendesk’s Suite plans come with a small number of included automated resolutions each month. As soon as you go over that limit, you start paying for every single extra resolution. And to get the really powerful AI features, you have to buy the "Advanced AI agents" add-on.

Here’s a quick look at the costs from their pricing page:

Plan Level	Included Automated Resolutions	Committed Pay-As-You-Go Price	Overage Price
Suite Team	5 per agent/month	$1.50 / resolution	$2.00 / resolution
Suite Professional	10 per agent/month	$1.50 / resolution	$2.00 / resolution
Suite Enterprise	15 per agent/month	$1.50 / resolution	$2.00 / resolution

With this setup, your costs are directly linked to how many tickets you get and how good your AI is at solving them.

The problem with paying per resolution

Here's the catch with a pay-per-resolution model: your costs can get unpredictable, fast. If you have a busy month with a spike in tickets, or if your AI agent actually gets better at its job, your bill goes up. You're essentially paying a penalty for being successful. This makes budgeting a nightmare and might make you think twice before letting your AI handle more complex issues.

It's a different way of thinking compared to a tool like eesel AI, where the pricing is straightforward. eesel’s plans are a flat monthly fee with plenty of AI interactions included. You don't have to stress about per-resolution charges or getting a surprise bill. It lets you scale up your automation with confidence, without your costs spinning out of control.

A visual of the eesel AI pricing page, which contrasts with Zendesk's model by showing clear, public-facing costs.

What really counts as a "resolution"?

Another tricky part is the fuzzy definition of a "resolution." Zendesk's own documentation says a resolution can be counted if a customer doesn't respond for 72 hours. But what if the customer just got annoyed and left? Or decided to call you instead?

This is a pretty common worry for Zendesk users, and it leads to a lack of trust in the numbers you're being billed for.

This is where eesel AI’s simulation mode is a huge help. Before you ever turn your AI agent on, you can run it on thousands of your past tickets in a totally risk-free environment. You can see exactly how it would have performed, what its resolution rate would have been, and what your ROI looks like, all based on your own real customer conversations. It takes the guesswork out of the equation and gives you the confidence to go live.

The eesel AI simulation feature provides a safe, risk-free environment to test AI performance on past tickets.

A better way to manage support

True success with AI isn't just about how many resolutions you can count. It’s about having fine-grained control, getting insights you can actually use, and working smoothly with the tools you already have. Here’s how eesel AI offers a more modern take on things.

Go live in minutes, not months

eesel AI is designed so you can set it up yourself, fast. You can connect it to your helpdesk with a single click, no messing with complex APIs or sitting through long sales demos. It fits right into your existing workflows instead of making you change how you work. While some big enterprise AI tools can take months to get going, you can have a powerful AI agent up and running with eesel in just a few minutes.

A flowchart outlining the quick, self-serve implementation of eesel AI, from connecting data to going live.

Test with confidence using risk-free simulation

Instead of just crossing your fingers and hoping for the best, eesel AI lets you test everything out first. The simulation mode gives you an accurate preview of your resolution rate and how much you could save.

This also lets you roll out your AI gradually. You can start by having it handle just one or two types of tickets, see how it does, and then expand its duties as you get more comfortable. You’re always in the driver's seat, making decisions based on how the AI performs with your customers, which takes all the risk out of the process.

Get actionable reporting

A good analytics dashboard should tell you why things are happening, not just what happened. eesel AI’s reporting is built to be useful. It doesn't just show you a resolution rate; it actively points out gaps in your knowledge base by telling you which common questions it couldn't answer.

This creates a feedback loop that helps you get better over time. eesel can even help you fill those knowledge gaps by automatically drafting new help center articles based on how your human agents successfully resolved similar issues. It gives you a clear path to improving both your AI and your self-service options.

A screenshot of the eesel AI reporting dashboard, which highlights knowledge gaps and other actionable insights beyond simple metrics.

Moving beyond simple metrics

So, while keeping an eye on your Zendesk AI agent metrics to track automated resolutions is important, it's pretty clear the model has its downsides. The per-resolution pricing can lead to surprise bills, and the fuzzy definitions can make you feel like you're not fully in control.

A modern AI tool should help you grow, not charge you more for it. By focusing on risk-free testing, predictable pricing, and useful insights, eesel AI offers a smarter way forward. It plugs directly into your Zendesk account, giving you the power to automate support effectively without all the risk and complexity.