Zendesk has introduced AI agents, baked right into their existing plans. It’s a big move that marks a continued evolution of AI in customer support. Being able to test Zendesk’s AI capabilities with your existing subscription is great, but beware the limitations on automatic ticket resolutions and extra costs. 

Zendesk bots and Zendesk AI agents are two terms that many use interchangeably, but they’re two very distinct features. We’re here to show you the difference and discuss some of the difficulties some users have had.

After reading this review, you’ll be able to create a bot, test it out with AI capabilities, and make an informed decision about whether it’s worth your time and money, or whether alternatives might be a better fit for you.

So, what’s the difference between Zendesk’s bots and Zendesk AI?

Chatbots are based on predetermined rules to help route the customer to a particular answer. In contrast, AI chatbots use NLP (Natural Language Processing) or machine learning to contextually understand a user’s query and formulate a response using a knowledge base provided to it. 

In the case of Zendesk, their bots are based on conversational flowcharts that support teams create to connect certain types of questions and keywords to predetermined responses by the bot. Zendesk has introduced Generative AI, which is a feature you can turn on for your bots that will be able to formulate responses based on your help center rather than relying on the predetermined ‘question and answer’ flowcharts.

Zendesk also offers Advanced AI which includes upgraded features like intelligent triage and macros insights, but this can be pricey. We’ll focus on Generative AI in this article so we can make sure we cover everything that’s offered in the base plans first.

Let’s start by creating our first bot automation and making sure we understand how it works, before we add any Generative AI functionality.

Creating a Zendesk Bot

The process of creating a Zendesk bot is easy - until it gets complicated. Below, we’ll demonstrate how to set up the basics and what other resources or processes you need in place so that it can work the way it’s supposed to.

Accessing the creation tool

To begin the bot creation process, you have to have access to and navigate through Zendesk's administrative interface:

1. Log in to the Zendesk Admin Center using appropriate credentials (you can’t create through the Agent Workspace).

2. In the main navigation menu, locate and click on "Conversation Bots." This section houses all bot-related tools and settings.

3. Within the Conversation Bots area, select "Create New Bot" to initiate the bot creation wizard.

   
   

Setting up the bot

It may seem overly simple, but this initial setup phase involves several key decisions that will shape the bot's basic functionality

• Naming the bot: Choose a name that reflects your brand and the bot's purpose. This name will be visible to customers, so consider its impact on user perception.

• Provide an avatar: something that represents your company and makes it clear your customer is speaking to AI.

• Saving initial settings: After making these selections, save the basic configuration. These settings can usually be adjusted later if needed.

  
  

Naming your bot Susan with a picture of a kind woman will probably throw people off if they realize after a few messages that they’re talking to an AI. Not exactly building trust in your brand, in that case. It’s much better to make it clear who and what customers are talking to from the get-go.

Defining bot interactions

Zendesk will provide some basic answers for you to create. You can then match multiple questions to each answer, so question variations can provide accurate results:

1. Creating questions and responses:

   • Develop a comprehensive list of frequently asked questions (FAQs)

   • Craft clear, concise responses for each FAQ. Make sure you link to relevant documentation when necessary.

   • Consider variations in how questions might be asked to improve matching

     
     

2. Implementing best practices for conversation flow:

The groundwork research for your bot will be very important in determining how your bot will be able to handle customers, both with and without generative AI. It’s important to establish what information you’re missing from key places like your help center before you can expect your bot to be able to answer anything for you.

• Understanding customer needs: Analyze customer service data to identify common issues and queries. You probably have a whole host of past tickets that will show you exactly the kind of queries your customers will be asking in the future.

• Outlining steps to meet customer goals: Break down complex processes into manageable steps, but don’t be afraid to escalate to an agent when needed.

• Designing answer flowcharts: Flowcharts in Zendesk will determine the exact process the bot will take when certain conditions are met. Think “if this, then do that”. The conversation flow charts can be a little daunting, but once you play around with it and follow the tips, it’s not as complicated as it looks.

• Prioritizing common use cases: Focus on high-volume inquiries first to maximize initial impact

4. Testing bot interactions:

• You can preview and test the bot in the admin center once it’s connected

• Conduct thorough testing with various user scenarios to make sure your conversation flows are well structured. It might be a good idea to ask someone else to test it for you so they don’t expect what the bot will or won’t know in advance!

• Include edge cases and unexpected user inputs to see how the bot will react.

• Refine responses and conversation flows based on test results

Generative AI 

Now that we’ve set up a bot automation and successfully created some conversation workflows, it’s time to check out generative AI. This features is available on all plans, but there are limited resolutions per month so it’s important to test it thoroughly before deploying.

What does generative AI do?

Zendesk's generative AI capability allows for more dynamic, context-aware responses than its bot automation counterpart. There is greater flexibility as the AI will draw information from knowledge sources and explain answers to more complex questions, rather than simply routing the user to the ‘best guess’ response. 

AI will usually also have a more conversational tone and is more likely to be able to handle unexpected phrasing or questions that are less common.

Unfortunately, Zendesk has limited the features you can experience on the lower payment plans. Their advanced AI features are onl available on professional and above, and is also a paid add-on.

If you’re ready to cough up some extra coin, however, you’ll be granted access to features like:

• Summarizing tickets

• Intelligent triage

• Intelligence in the context panel

• Generative AI for agents

• Macro suggestions for admins

• Autoreply and internal note trigger actions

• Generative AI for help center

• Generative AI for voice

We’re just covering the basics of what you get without the add-on, but if you want more resources, check out our guide on Zendesk Advanced AI here. Or, if you’re confused about all the pricing details, you can have a read of our Zendesk pricing guide here.

How to set up generative AI

This is the easy part. You can either select it when you first set up the basics of your bot as see below:

Or, you can turn it on for an existing bot by selecting the option to generate replies under the behaviour section of the bot settings.

Creating bot personas

Bot personas are a way to determine how your AI will respond to messages, making sure the expression and tone they use is the one you want your customers to associate with your brand. There are a few pre-developed personas to choose from, which can include emojis. However, the choices and customization are relatively limited.

You can turn on and select bot personas in the Admin center, under channels → bots and automation → bots. From there, select the bot, and select personas.

Optimizing your bot

Your bot won’t be able to intuitively understand where to get your information from. Depending on what you’re using it for, you need to provide it with access to the right knowledge base. It’s also important to fill that knowledge base with accurate, clear, and comprehensive answers to anything your customers might need to know.

Preparing help center content

A well-structured knowledge base is essential for effective bot operation. Here are some steps you can take to make sure your AI agent has all the tools it needs to do its job right.

• Review existing content: Ensure all articles are up-to-date and accurate. This can be difficult if content is spread across multiple sources.

• Update information: Reflect any recent changes in products, services, or policies

• Structure for accessibility: Organize content logically for easy bot retrieval - if you can’t find it, neither can your bot.

• Use clear, consistent terminology: Align knowledge base language with bot responses so that it can match the persona you’re asking of it.

• Implement proper tagging and categorization: This helps the bot quickly locate relevant information

Common Complaints and Issues

Despite their benefits, Zendesk AI bots are far from perfect. We’ve put together a few common complaints that some people have recognized so far. These issues aren’t exclusive to Zendesk’s AI, but it’s important to remember that if you’re unable to get value from their AI features you should look for alternatives that may have already overcome these limitations. 

Pricing concerns

Zendesk is an extremely popular platform and because of this they have the freedom to charge for things that many believe they shouldn’t. Below you can find the pricing plan for their suite plans, which all include Generative AI. 

Offering AI built into their existing plans is a very attractive marketing campaign. But it doesn’t match what they’re really doing. Their built-in generative AI is extremely limited and only touches the tip of the iceberg of what their AI is capable of. In order to access the rest, you not only need to be on a professional tier or higher, you also need to pay $50 per agent, per month. 

For a company with 10 agents, paying annually, and who want to include the Advanced AI add-on in order to triage tickets and utilize macro insights, they would need to at least be on the Suite Professional Plan. This would be $19,800 per year. Adding just one more agent to your company would increase this cost by almost $2,000. This does not include the extra expense of automated ticket resolutions, which have pricing issues of their own.

As you can see in the comments of some of their community support forum posts, some users are finding the decision making behind automated resolution pricing to be dubious.

One Zendesk user says: "A resolved conversation is when the bot asks if the issue is resolved yes/no? If the client confirms that yes, it is resolved. Then it is resolved. If a client just leaves a conversation, it doesn't mean that it is resolved. I'm surprised that you're going to charge $1.5 per such conversation."

(Source: Zendesk Product Guide Article "About automated resolutions for AI agents.")

This doesn’t fly well with a lot of customers, and can quickly rack up the expense, especially considering the add on doesn’t increase your resolution usage allowance either.

Mismatched responses

Many users have reported issues with conversational or ‘small talk’, or simply struggle with recognising when the generative AI responses should be triggered. Instead they simply route to predetermined answers of ‘sorry’, and routing to an agent instead. These mismatched responses can also be:

• Intent recognition failures: Bots may misinterpret user queries, especially with complex or ambiguous language

• Irrelevant information provision: This can frustrate users and break down trust in the automated system and the brand itself

Limited functionality

Bots sometimes struggle with multi-step processes or queries requiring nuanced understanding. Without the Advanced AI addon, agents taking on board these queries won’t be benefited by Generative AI. The basic AI features only allow the bot to draw from the help center and don’t provide enough functionality to have a meaningful impact on user queries that don’t fall within certain parameters.

Language and localization issues

Several Zendesk users have reported issues with the bot responding with different languages, and having limited languages available to select. 

In addition, there have been several instances of misinterpretation or miscommunication of the following:

• Accurate translation of industry-specific terms

• Understanding regional dialects or colloquialisms

• Maintaining consistent tone across languages

• Vulgar language being used in a funny but detrimental mistranslation error.

Integration problems

Zendesk bots and AI are able to integrate with multiple channels, but they lack the ability to pull information from sources outside of their help center. 

If your documentation is spread across multiple platforms, it might be a struggle to make sure all your information is up to date and accessible on the help center. 

And if you’re going to change help desk platforms in the future, the bot, AI customization, and your help center documentation won’t be coming with you.

What’s the alternative?

The good news is that Zendesk is not a standalone platform. They support a great many apps that you can integrate to improve you, your agents’, and your customers’ experience. This includes, but is not limited to, AI bot integrations that can replace or help expand Zendesk’s native offerings.

We’ve listed out a few Zendesk AI alternatives here, which should provide you a brief overview of what’s out there. 

Included in this list is of course our very own eesel AI, which addresses a lot of the concerns that many users have had with Zendesk’s bot and AI offerings. We’re proud to be able to link with multiple knowledge sources like Google Drive, or Confluence, and provide the same features that Zendesk does, and more, for a much more reasonable pricing plan.

Summary

Zendesk is a popular and powerful platform for your help desk needs. Their bot automation and AI offerings promise a great deal of productivity efficiencies for your agents. 

Setting up a bot and starting to use Generative AI is an easy process. But optimizing it and making the most of the limited features you get without the addon is difficult. 

The Zendesk AI addon provides a lot of features that eesel AI provides, but with a hefty price tag that will only increase as your business scales. 

Of course, every business is different and so is their customer base. For some smaller businesses, the conversational personas and drawing information directly from the help center might be the perfect amount of AI automation to really boost their customer support.

However, when considering the many issues Zendesk users have faced when attempting to implement AI, in combination with the pricing issues, it seems Zendesk has a long way to go. 

Hopefully this setup and review has been able to help you test out your own bots and Generative AI, so that you can make an informed and measured decision about the direction of your customer support.