Zendesk AI: Leveling up with generative AI for a more intuitive and intelligent CX platform

Zendesk is expanding the use of artificial intelligence (AI) across its customer experience (CX) platform with the release today of a series of new capabilities the company is branding as Zendesk AI.

The new Zendesk AI capabilities include advanced bots for handling customer inquiries across a range of industries. Customer service agents will now also benefit from AI-powered assistance in responding to inquiries. Additionally, the system integrates intelligent triage capabilities that use sentiment analysis and intent detection to help route requests properly. 

While many of the Zendesk AI capabilities have been built by the company’s own teams, the new services have also benefited from a healthy dose of generative AI via a partnership with OpenAI. The generative AI component will fit in across the Zendesk AI suite, helping to generate responses and summarize content.

>>Follow VentureBeat’s ongoing generative AI coverage<<

“We believe we need to make AI easy to understand and use and accessible to everyone,” Cristina Fonseca, head of AI at Zendesk, told VentureBeat.

What generative AI adds to Zendesk

Zendesk was no stranger to AI prior to its partnership with OpenAI.  

Zendesk took a big step into AI in 2021 with the the acquisition of Portugal-based startup Cleverly AI, where Fonseca was formerly the CEO. In September 2022, Zendesk announced a major update integrating the Cleverly AI technology to help speed up customer service responses. Fonseca said today’s Zendesk AI release is an evolution of what the company announced in September 2022, which was targeted specifically at the retail industry. She noted that the new release is available to a broader group of customers.

Zendesk has offered bot capabilities for some time, but Fonseca said OpenAI’s generative AI foundation has significantly expanded those capabilities.

“We believe software should be intelligent off the shelf and, for most bots to work, you need to spend a lot of time training them,” she said. “This is one of the main features we are now launching — with bots that are pretrained and already understand customers.”

The generative AI will also help power a revamped set of capabilities for increased customer service agent productivity. Zendesk AI uses OpenAI’s technology to support summation and sentiment analysis for inquiries that, in turn, help agents respond effectively. The ability to assist agents with creating replies is also part of the new update.

Why generative AI alone isn’t enough for CX

Fonseca emphasized that while Zendesk AI is making use of generative AI from OpenAI, it is not abandoning its Cleverly AI roots.

Intelligent triage, for example, is a foundational element the product uses that was developed by Cleverly AI for understanding customer intent, sentiment and language to appropriately direct an inquiry. For that system, Zendesk has its own proprietary models that were trained on Zendesk’s data. Fonseca said those models understand customer service because they were specifically trained on customer service data and provide a high degree of accuracy.

“The way we see [OpenAI’s generative AI] is basically as a tool to help us accelerate things that were already in the roadmap, and make a ton of sense to add to our suite of products,” Fonseca said. 

As an example, she noted that Zendesk was building out its own approach to suggesting new replies for customer service agents, as well as creating content for a knowledge base about different issues. Without OpenAI, she said the draft content that Zendesk was able to generate was not quite polished. Now with the OpenAI integration, the data from Zendesk can be used to generate well-written replies and knowledge base articles.

“We are trying to leverage OpenAI and large language models (LLMs) to help us perfect everything we do, on top of content,” she said.

With all the power that AI brings to Zendesk and its users, Fonseca cautioned that it’s important to realize that AI can’t and shouldn’t do everything when it comes to customer experience. She noted that CX is complex and it includes workflows — and sometimes systems — that are not integrated with Zendesk.

“I think the first thing we should do for our customers is help them understand what should be automated and what shouldn’t be automated,” she said. “Because if something cannot be automated, there’s no point in you trying to have a bot talk to a customer if the bot cannot add any value.”