top of page
  • Writer's pictureAnne-Marie Traas

Expert tips for creating chatbots with Zendesk

zendesk chatbot

In today's digital world, businesses are constantly on the hunt for the next big thing in customer service efficiency — and many are finding their answer in the dynamic world of chatbots. 


When integrated with Zendesk, chatbots aren't just helpful; they're transformative. They provide instant responses and simplify the entire support process. 


Whether you're just starting out with a small venture or managing a sprawling corporate empire, embracing a chatbot could propel your customer service to new heights. 


Let’s explore some practical tips for crafting, launching, and fine-tuning your very own Zendesk chatbot to captivate your customers and streamline your operations. For more high-level guidance on building chatbots, check out our blog post on CX best practices for chatbots and self-service.


Start with clear objectives


What exactly do you want your bot to achieve?


Your bot can handle so many tasks and fit into so many processes that it may take time to bring about its full realization. A common use case is for a bot to be the first point of contact for website visitors who want to contact you, including customers who need support and leads who are interested in your services.


If your main goal is to deflect tickets and reduce the workload of your agents, you’ll want your bot to address FAQs with answers and useful help articles from your knowledge base.


To help convert leads, your bot can collect their information before connecting them with a sales agent or scheduling a demo for them. Having clearly defined objectives will help you establish just how much content your bot will need.


Zendesk has written a very comprehensive article giving some suggestions for how their bots can be utilized in your processes - I recommend adding it to your reading list!


Optimize language for clarity and tone


Your chatbot may be an automated tool, but it should still sound like a part of your company to create a more welcoming image for your brand. From start to finish, the conversation should feel personal and more human than robotic.


Keep your conversations brief and to the point to avoid overwhelming your customers. By focusing on clear, concise communication, you'll help resolve their issues more quickly and create a more pleasant interaction, enhancing overall satisfaction. 


Tone is also key in creating your chatbot’s persona. You want your bot to feel like an extension of the team and not like…well…a bot. Tailor the responses to sound like something your support team would say. If you aim for friendly and informal interactions from your agents, your bot should follow the same guidelines.


Apply these tips when formulating your responses:

  • Keep it simple. Use plain language that is easy to understand. Avoid using jargon or industry-specific phrases that may confuse somebody who isn’t an expert.

  • Be direct. Get straight to the point. Break down information into shorter, manageable sentences. This makes the content easier to read and understand quickly, which is crucial for maintaining the user’s attention and ensuring clarity. Direct the user to a help center article for more in depth reading.

  • Focus on one idea per message. Each message from the chatbot should focus on a single idea or topic to avoid overwhelming the user. 

  • Incorporate natural language. Write your responses as if you were having a conversation with your user. Show empathy to make the interaction more personal. If a user is reporting a bug or outage, acknowledge that they may be feeling a little annoyed: “Oof, that sounds frustrating! Let me help you with that.”

  • Give your bot a (human) name. Human names are more personable than something like Super Automaton 3000. If you have a company mascot, name your bot after her. Alternatively, choose a member of your support team to be the name and face of it. If all else fails, a generic name like John or Kim also works.

  • But don’t be afraid to admit it’s a bot. Remember in the ‘90’s when it was all the rage to fool callers into thinking you’d answered when it was actually your answering machine? It may have been funny for you, but it was annoying for the caller who got tricked. Don’t replicate that experience with your chatbot.  


Use conditional logic for smarter interactions


Zendesk recently added Generative AI to its chatbot offering, making it possible for a chatbot to have a fairly open-ended and human-like conversation. A customer can freely type what they need help with, and the bot will respond with relevant answers pulled from your knowledge base.


This takes some significant effort to apply, though. You need to teach the AI all about your product, as well as all the different ways it could be asked about it. If you have a thorough and accurate knowledge base, it’s a lot easier. But if you have features that could be confused for each other — like an iPod and an iPad — you risk the AI providing incorrect information.


While generative AI is great, sometimes it’s still helpful to give the customer options to choose from. Each option produces more choices until you either reach a resolution for the customer or route them to an agent. Here’s an example of a chatbot workflow:


zendesk chatbot flow

Don’t get carried away and overwhelm the user with choices, though. 


Limit them to a handful of general options, with each one branching into a more specific topic related to the previous choice. Try to wrap up the interaction within 4 or 5 steps so your customer doesn’t become frustrated. While Zendesk allows up to 2,000 steps and 500 answers in your chatbot flow, don’t feel like you need to use all of them!


GenAI is great for the occasions where you haven’t programmed a specific answer to a question. Zendesk’s generative replies AI is smart enough to search your knowledge base for information, condense it into a concise answer, and still maintain your company preferred voice and tone. It can bolster the process and help fill in the gaps that might be left out of your chat tree.


Generative AI more than justifies an article all by itself rather than being relegated to a mere blip of a chapter in this one. Luckily for us, our friends at Zendesk already wrote one. They take a deep dive into the different forms their GenAI can take, and how each can be used in CX.


Regularly test and iterate your Zendesk chatbot


Like all of your processes, Super Automaton 3000 will need tweaks along the way. Following the continuous improvement cycle, you’ll test, review feedback, and refine to get as close to perfection as possible. 


For testing, get someone outside your support department to use the chatbot. Are they able to get to the answer they are looking for? Do the interactions feel on-brand? Monitor real-time interactions with your customers to see how they are faring. Are they giving up, or having to backtrack frequently? This could be a sign that your presented options are unclear.


Feedback comes in a number of ways. Maybe a frustrated user has submitted a ticket titled “I hate your robot. Why can’t I just talk to a human?” That’s pretty direct feedback right there, though not very useful. Ask them for more information. Do they just hate all things AI as a general principal, or do they have something more constructive to add about your specific bot?


Analytical information is another valuable form of feedback and is provided by Zendesk in the form of Zendesk Explore. It allows you to monitor how many conversations the bot has over a given time period, how many were resolved or escalated to a human, and how many were abandoned. You can also see exactly which branches of your flow have been followed.


From this information, you can identify areas of your bot that aren’t working. If abandonments are happening frequently on a particular branch, check the wording of the messages and that the information being provided is correct and clear. If a customer gets seven steps through the branch and gives up, it’s probably too long.


Integrate with other Zendesk tools and apps


There are a number of functions and apps from Zendesk and external providers like Swifteq that can further elevate your chatbot and make the customer experience more seamless:

  • Zendesk Guide. You (should) already have a knowledge base. It’s time to put it to further use. Link articles directly from your help center to provide more comprehensive answers without the need for lengthy messages. A help center powered by Swifteq’s Help Center Manager will already be well organized, making relevant articles easy to find.

  • Zendesk Explore. This tool can help you analyze conversation patterns, identify common issues or questions, and refine your responses or decision trees based on real usage data.

  • Zendesk Support Tickets. Integration with the support ticket system allows the chatbot to create, update, and retrieve ticket information, providing a seamless transition from chatbot to human agent when necessary. Automate routine tasks and ensure a continuous, connected experience for the user. 

  • Swifteq's Triggers+. Swifteq takes what Zendesk already offers and adds a significant dose of steroids to it. Quickly extract information from a chatbot conversation, create tickets, start live chats, or even bring ChatGPT into the equation. The possibilities are almost endless. 


These tips should help you get started with your chatbot. Experiment to find what works best for your company. Make it sound like your support department; use simple terminology as often as possible, while making messaging sound conversational rather than procedural; make the process as efficient as possible by providing logical paths to follow; keep reviewing and testing, using feedback and analytics to inform your improvements; exploit integrations to make your bot as powerful as possible.


Zendesk already provides a great platform for building your bot, while Swifteq gives you the tools to transform it from a functional bot, providing simple answers and gathering information, into a fully functional member of your support team. Setup a demo with Swifteq to see which other tools can assist you along the way.



Anne-Marie is a customer success executive focused on communications and scalability. She specializes in driving process & product improvements, creating thorough and easy-to-understand product documentation, and teaching others how to communicate more effectively through the written word. You can find her on LinkedIn



Comments

Couldn’t Load Comments
It looks like there was a technical problem. Try reconnecting or refreshing the page.
Start your week with great quality articles on customer support

Thanks for submitting!

Subscribe
bottom of page