There are two approaches to automation. One is quick and inexpensive (at first), allowing you to use a set of out-of-the box intents from benchmark data that may or may not be specific to your industry. It’s a step up from relying on basic, rules-based chatbots, and it’s great if you don’t want to rely too much on automation, or are just getting started.
The other relies on historical data from your very own past live chat or email support interactions. The benefits? A virtual agent that offers a hyper-personalized customer experience, an AI model that grows more precise with every new customer interaction, and an ideal allocation of your support team’s time and costs.
It’s option two that’s what you’ll want for the long haul, and Ultimate’s Zendesk integration will help you get there in minutes.
Import your historical data (aka your customers’ most pressing inquiries)
First, import your historical data using a Zendesk authorization. You can use chat transcripts or email exchanges for this, with absolutely no need to worry about data security, as both Zendesk and Ultimate are GDPR compliant and undergo regular security audits, such as SSO and SOC2.
Train your conversational AI model
Now, it’s time to work with what you have, and analyze your data using our CS Automation Explorer. This no-code tool helps generate insights into automation potential and prioritize common topics for your next steps. It also sets the stage for the training phase, which is when your AI model starts matching intents with expressions.
As a rule of thumb, aim for 50 to 100 expressions per intent for your most common intents. For example, your customers may come up with tons of ways to ask for the status of their order. They may wonder, “Why hasn’t my package arrived?”, demand, “Hey, can you send me a tracking number for my order?”, or want to know, “Where’s my new pair of Jordans?”.
Build your dialogues and enhance your chat UX
This is the fun part! You can now plan conversations to help solve the intents you’ve discovered in your CS Automation Explorer. Begin by carefully mapping out your conversations between your virtual agent and your customers, and thinking about the different turns they might take. This includes deciding which UX features to use — buttons, free text fields, or rich media are creative and convenient ways to offer and collect information in a dialogue flow.
“Our Zendesk integration works seamlessly within our system and can create amazing customization without any additional web development work by the customers.”
- Victoria Lojek, our Senior Solution Architect
Let your Zendesk integration shine
This is the moment where everything falls into place. In your Zendesk settings, you can now trigger global bot actions and personalize your conversational flows using the dialogue creator dashboard.
Make sure to test your dialogues before going live. A bullet-proof way to do this is by hosting a “break your bot” session, where you invite your team to interact with your virtual agent and ask it as many questions as possible to pinpoint questions you may not have anticipated while building your dialogues. Even better? It’s part of our 14-day Ultimate Bootcamp onboarding process, so rest assured you are in good (in fact, the Ultimate) hands.
Launch your virtual agent
In the weeks and months after going live, it’s important to observe, refine, and adjust your AI model. This step is now easier and more effective than ever thanks to Ultimate’s new Training Center. It allows you to harness our deep AI, built in-house, to make your virtual agent grow even more accurate with every interaction.
“There are many things I like about this tool: it sits in our existing Zendesk support system, it is fun to train and work with, and now the bot is the team’s best friend!”
- Philipp Weidenbach, Head of Patient Operations, TeleClinic
Find out how Teleclinic, a German telemedicine platform, were able to serve their customers across email and chat by integrating Ultimate into Zendesk, after their Customer Service team grew by 150% during the pandemic.