How to use Zendesk Explore recipes to measure Answer Bot resolutions

So, you’ve got Zendesk’s AI agent (the tool formerly known as Answer Bot) up and running. You're hoping it will handle some of the more common questions and give your support team a bit of breathing room. That’s a solid move.

But now for the real question: is it actually working? How many tickets is it really resolving on its own? Do customers find the article suggestions helpful, or are they just getting frustrated and waiting for a human anyway? Getting clear answers here is the only way to know if your investment is paying off and how to fine-tune your self-service strategy.

Lots of folks head straight to Zendesk Explore to build custom reports, but let's be honest, wrestling with datasets, metrics, and attributes can feel like you need a data science degree. This guide is here to help. We'll walk through a practical way to use Zendesk Explore recipes to measure Answer Bot resolutions. We’ll also get real about the common roadblocks you'll hit and show you a much simpler path to getting insights you can actually use.

What you’ll need to get started

Before we jump in, let’s make sure you have everything lined up. You won't get very far without the right plan and permissions.

• A Zendesk Suite Plan: You'll need to be on a Professional, Enterprise, or Enterprise Plus plan.

• Zendesk Explore: Make sure you have access to Explore Professional or Enterprise, as this is where the custom report building happens.

• The right permissions: You’ll need to be an Editor or Admin in your Zendesk account to create and mess with Explore reports.

• An active AI Agent: This might seem obvious, but your Zendesk AI agent needs to be enabled and actively offering up articles to your customers.

A step-by-step guide to measuring Answer Bot resolutions

Zendesk offers a few "recipes" to get you started, and one of the most useful is the resolution funnel report. It’s designed to show you the whole journey: how many times an article was suggested, offered, clicked, and finally, led to a ticket being solved.

Here’s how to put it together, one step at a time.

Step 1: Create a new report in Explore

First things first, you have to get into Explore and pick the right dataset. This part is really important, because choosing the wrong one means you won't have the data we need.

1. From your main Zendesk dashboard, find and click on the Explore section (it's the little graph icon).

2. In the Reports library, hit New report.

3. You'll see a "Choose a dataset" page. Find and select Answer Bot > Answer Bot - Article Recommendations, then click Start report.

Step 2: Add your key metrics

Okay, now we'll add the numbers that will build the stages of our funnel. We want to see the whole story, from the first attempt to the final resolution.

1. Look for the Metrics panel on the right and click Add.

2. From the list that pops up, you’ll want to find and select these four metrics under the Answer Bot answers section:

   • Attempts: This counts every time the bot had a chance to suggest an article.

   • Answers: This is when the bot actually found and offered at least one article.

   • Clicks: This tells you a customer clicked on one of the articles the bot suggested.

   • Resolutions: This is the big one, it's when a customer marked their ticket as solved after clicking a suggestion.

3. Click Apply.

Step 3: Visualize the data as a funnel

A list of numbers is fine, but it doesn't really tell a story. Let’s switch it to a funnel chart so we can see exactly where people are dropping off.

1. Click the Visualization type icon on the right-hand menu.

2. Select Funnel.

3. Next, click the Chart configuration icon (it looks like a little paint palette) and go to the Chart tab.

4. Change the Computation method to Percent of the first element. This handy setting shows each stage as a percentage of the total "Attempts," which makes it way easier to see your conversion rates at a glance.

And there you have it. You've built a basic funnel report that shows how well your Answer Bot is turning questions into solved tickets. You can play around with filters (like "Answer created - Date") to see how things look over different weeks or months.

Common headaches with Zendesk Explore recipes

While that recipe gives you a decent starting point, most support managers find themselves hitting a wall pretty quickly when they try to dig any deeper. The built-in Zendesk reporting is powerful, for sure, but it comes with some serious baggage and blind spots.

• It’s a ton of manual work. Building, tweaking, and just keeping these custom reports alive takes some serious know-how. It's not a "five-minute" task, and it pulls you away from what you should be doing, like coaching your team. If you need proof, just look at the dozens of official Explore recipes Zendesk has published. It’s a pretty clear sign that getting the answers you need is a complicated, manual process.

• Your data is all over the place. Zendesk keeps its bot analytics in different buckets, like "Article Recommendations" and "Flow Builder." This makes it a real pain to get a single, clear picture of what’s going on. You might see how your articles are doing, but it’s nearly impossible to connect that back to the conversation that got the customer there in the first place.

• "Resolution" doesn't mean what you think it means. The default reports mostly track a resolution when a customer clicks an article and then manually closes their ticket. It has a tough time spotting resolutions that happen purely through conversation, or when the AI helps someone without them having to read a long article.

• You get numbers, not answers. The funnel chart is great for showing you what happened, but it doesn't tell you why. Why are so many people not clicking the suggestions? What topics are your help articles completely missing? Zendesk Explore won't point out these knowledge gaps, so you’re left guessing what content you need to create or fix next.

Relying only on the native tools can really slow you down. You end up spending more time trying to build reports than you do acting on them.

A better way to measure Answer Bot resolutions

Instead of losing hours wrestling with datasets, what if you could just plug a specialized AI platform directly into your Zendesk account? This is where a tool like eesel AI changes the game. It seamlessly integrates with Zendesk and gives you the deep, useful reporting you need, right out of the box.

Finally see your real resolution rate

eesel AI does more than just count article clicks. It actually understands the context of a conversation. It learns from all your past tickets, macros, and can even connect to other knowledge sources you use, like Confluence or Google Docs. Because of this, it can measure true automated resolutions with much greater accuracy.

The analytics dashboard gives you a clean, simple view of:

• Automated Resolution Rate: The percentage of chats and emails that the AI handled completely, with no human agent ever needing to step in.

• Cost Savings: A straightforward calculation of how much money you’re saving, based on your average cost per ticket.

• Knowledge Gaps: The AI points out the top questions it couldn't answer, basically handing you a to-do list for what new help content to create.

The eesel AI dashboard provides clear insights into knowledge gaps and resolution rates, offering a better way to measure performance than standard Zendesk Explore recipes.

Test your setup before it goes live

One of the scariest parts of rolling out any AI is the uncertainty. You're left wondering, "How is this actually going to perform with my customers' weirdly specific questions?"

With eesel AI's simulation mode, you can test your AI on thousands of your own historical Zendesk tickets before a single customer ever talks to it. It gives you a surprisingly accurate forecast of your potential resolution rate and shows you exactly how the AI would have answered real questions from your customers. This takes all the guesswork out of the equation and lets you go live with confidence, something you just can't get with the standard Zendesk tools.

eesel AI's simulation mode allows you to test your setup on historical tickets before going live, a feature not available when using only Zendesk Explore recipes to measure Answer Bot resolutions.

Pro tips for measuring Answer Bot resolutions

Whether you stick with the native tools or bring in a platform like eesel AI, here are a few things to keep in mind:

• Don't get fooled by "deflection." A "deflected" ticket isn't always a win. If the customer leaves frustrated, you haven't really helped. Focus on the actual resolution rate and customer satisfaction to make sure your bot is being helpful, not just a roadblock.

• Treat unanswered questions like gold. No AI is perfect. When your bot can't answer something, don't see it as a failure. See it as a gift. Those questions are a direct line into your customers' biggest pain points and your single best opportunity to improve your knowledge base.

• Start small, then build. Don't try to automate your entire support operation overnight. Use a tool that lets you roll out automation to specific channels or for certain types of tickets first. See how it goes, learn from the results, and expand from there as you get more comfortable and your knowledge base gets stronger.

Stop building reports, start getting answers

Figuring out how your support AI is performing shouldn't feel like a second job. While Zendesk Explore recipes can give you a peek into your Answer Bot resolutions, they often create more work and leave you with more questions than answers.

By using a platform built for the job, like eesel AI for Zendesk, you can skip all the tedious report-building. You get to jump straight to the insights that will actually help you lower costs, beef up your knowledge base, and give your customers a better experience.

Curious what your true resolution rate could be? Sign up for eesel AI and run a free simulation on your Zendesk tickets in minutes.