Customer Support Academy

Enterprise Bot Building Best Practices

In 2023, it’s no longer a matter of if you should incorporate a bot into your CX team, but when. This guide offers expert tips and actionable advice from Zendesk’s bot building team that will ensure you’re set up for success.  

Start module

By the end of this module you’ll learn:

  • Why your CX team is an essential first point of contact when deciding what to automate
  • Who the key players are on a successful bot building team and how to effectively liaise between them
  • How to set a strong foundation in your bot building flows that will create a consistently high quality customer experience
  • What to pay attention to as you monitor customer feedback data from conversations with your bot
  • Which automation tools will enable you to make quick changes to your bot when the unexpected happens
20 min. Clock

Meet the experts:


Austin Lacey

Manager, Chatbot and Self-Service Automation at Zendesk


Iulia Lipatov

Chatbot CX Designer at Zendesk


Hosam Hassan

Product Marketing Manager at Ultimate


Ehsan Ali

Automation Expert at Ultimate


Marika Sarén

Lead Automation Consultant at Ultimate

Enterprise bot building in 2023

In today's customer service landscape, where instant communication and round-the-clock availability are paramount, enterprise chatbots have emerged as powerful allies for businesses across the globe. And with new technological innovations like generative AI — the tech powering ChatGPT — the quality of these automated interactions is rapidly improving. Accordingly, our 2023 trends survey found that 92% of respondents say their trust in AI has increased over the past 12 months. What’s more, 48% of businesses are actually creating automation-related roles in their CS departments. As is often the case, however, enterprise companies may understandably have concerns over how complex it is to get started, how secure automation platforms are (tip: look for one that’s SOC2 compliant), and if a solution will be flexible enough to meet their needs as they scale. 

But if you’re reading this, it’s likely because you are already compelled by the case for automation and its ability to mitigate long wait times, make your support agents’ lives easier, and improve the overall customer journey. That’s why we’ve joined forces with our valued partners at Zendesk, who have been championing digital customer service since 2007 and who today connect more than 100,000 brands with hundreds of millions of customers all over the world. This module is designed to help you figure out where to start and what to expect once you get your bot up and running. In it, we cover key considerations from pre-launch, to those critical first weeks, as well as best practices for sustaining a successful bot over time.

"We find that our enterprise customers have some of the highest solve rates with our bot. Even though they have complicated setups, they still have questions that can be automated."

Austin Lacey
Manager, Chatbot and Self-Service Automation, Zendesk

1. Start small

Before you even begin your bot building endeavors, it’s worth defining the desired scope of automation and what your ideal results would be. Once you’ve established which use cases you are trying to solve, you’ll be able to make more impactful decisions about where to start automating. 

In this video on first steps for getting started, Austin Lacey, Chatbot and Self-Service Automation Manager at Zendesk, recommends that you don't begin with automating your meatiest, most complex topic. Work on some of the more granular elements while you get a hang of how to automate, establish a conversation style, and build out some initial bot flows. Tackle a few of your most repetitive FAQs such as order status or password change requests. From there, work on perfecting one channel at a time before scaling to omnichannel automation. 

"Start with a single channel and then expand from there. This kind of crawl, walk, run approach will help ensure that you set a benchmark for bot experience for your team to follow – something that will pay off in the long term. "

Ehsan Ali
Automation Expert , Ultimate

Start small to master the basics:



Before expanding to other channels, pick your highest volume channel to automate first.



Start with your most popular language before moving onto more.


Bot hours

Begin with a 9-5 timeframe where the bot is active, and then move to 24/7.


Brand tone & phrasing

Take this time to set standards on the conversation style of your bot.

2. Utilize your CX team

Once you’ve mastered the basics by testing out a few automation flows, it’s time to get strategic about how to make the biggest impact through your bot building efforts. A good starting point for understanding your customers’ needs is to utilize your CX team’s expertise. They are the ones closest to customers, so they know their pain points and their common questions. Iulia Lipatov, Chatbot CX Designer at Zendesk, emphasizes that your support agents offer the best touchpoint for choosing what to automate first. This could include simple things like adding tags in the CRM, routing tickets, and flagging topics of user questions if a query is handed over from bot to human agent. These interventions alone can have a huge impact on cutting agents’ workload so that they can focus on more complex tasks. 

“I am realizing how all my years working with customers helped me understand what they need and to build a customer focused mentality. Those skills really benefited positively once I joined the bot team.”

Iulia Lipatov
Chatbot CX Designer, Zendesk

3. Create smart escalation paths

Even if you’ve perfectly followed the first two bot building best practices, it is likely that something will eventually go wrong. For example, you may think you have a pretty clear idea of what the conversation is going to look like, but the user journey might be totally opposite. Here, Austin suggests that when mapping out all possible user flows – including what the conversation will look like if there’s a problem and a support agent needs to get involved. In this video on keeping a customer-focused mentality when it comes to bot building, Iulia reminds us that the goal should not be deflection at all costs. Rather, the bot should be treated as a tool that helps create a better customer experience.

"Keep in mind that when a chat is escalated to a human agent, it's an opportunity to improve your dialogue design. Editing just one block of content in your bot flow can lead to massive results."

Marika Sarén
Lead Automation Consultant, Ultimate

4. Assemble a diverse team

Your bot is only going to be as high performing as the team maintaining it, and both Iulia and Austin stress the importance of building relationships across functions. This can look like connecting CX knowledge with conversation designers as well as learning the lingo necessary to communicate with engineers and IT managers/developers who will help integrate the bot into your company’s existing internal systems. These kinds of project management skills centered on liaising between different functions are what’s going to keep your bot running up to speed and help you to quickly adapt in step with any changes or issues that might come your way.

EduHub-BestPractices_Step2_6 (1)

Assemble the ultimate project team:


Project lead

The project lead owns the project, and is responsible for internal success & communications.


Bot builders

Your team should include 2-3 bot builders, often with previous experience as support agents. They will own the bot building as well as monitor the chat content once it's live.


CRM admin

The CRM admin owns the CRM, widget, and technical setup.


Backend engineer

The backend engineer will be responsible for implementing backend integrations.

“Bot building requires so many different skills, and you can build entire teams of different specialists drawing on their unique backgrounds and expertise.”

Hosam Hassan
Product Marketing Manager, Ultimate

5. Don’t set and forget

The only way to make sure your bot really works is to launch it and keep iterating post monitoring customer feedback. In other words, it’s absolutely essential that you make data-driven decisions when it comes to optimizing your existing conversation flows. In this video about best practices for data monitoring, Austin emphasizes the importance of investing the time and effort in tracking conversations with customers to identify trends and apply customer feedback when tweaking your flows. 

If, for instance, a certain query is consistently escalated from your bot to an agent, it can offer a good opportunity to introduce a new API or make tweaks to the existing flows where clarity is needed. It is crucial to be able to pivot and adjust to these tendencies and trends in order to deliver the best bot experience and minimize your skip self-service rate among enterprise customers.


Key KPIs to consider when monitoring your bot flows:


First contact resolution (FCR)

FCR tracks the proportion of queries that are resolved in the customer’s first interaction with customer support.


Average handle time (AHT)

AHT measures how long it takes to resolve an issue. This metric is used to monitor a support team’s efficiency.


Drop off rate (DOR)

DOR refers to the rate at which customers do not complete their support queries. This metric helps you identify which points in the bot flow are causing customers to lose engagement.


Customer Satisfaction (CSAT)

Measured through customer feedback surveys, CSAT assesses how a product or service fares in comparison to a customer’s expectations.

“In addition to adjusting to customer feedback you need to be monitoring and watching for that feedback. Don’t discount the value of that monitoring all those conversations and to look for those trends.”

Austin Lacey
Manager, Chatbot and Self-Service Automation, Zendesk

Keep learning

Browse the Customer Support Academy for more courses on how to improve customer service