Want smarter insights in your inbox? Sign up for our weekly newsletters to get only what matters to enterprise AI, data, and security leaders. Subscribe Now
Zendesk’s customer experience platform is getting a fresh injection of artificial intelligence. The company is expanding its Zendesk AI solution by launching what it describes as “highly sophisticated” AI agents, an agent copilot and new capabilities that ensure proper support staffing and the quality of AI responses and case resolution. All these features constitute the end-to-end solution Zendesk believes brands want to manage the quantity and quality of customer interactions.
Between 10-20% of support interactions today are believed to be automated. An overwhelming majority still require a human to resolve. With AI innovation growing rapidly, companies are banking on the technology to tackle more support tickets while helping save money.
“In the past, we started with rules and workflows. We then went to machine learning [and] predictive analytics. And now, [generative] AI, large language models and the effectiveness of AI agents are massively better than…two or three years ago,” Zendesk CEO Tom Eggemeier said in an interview with VentureBeat. “We’ve been in the space for a while, but we’re seeing fantastic automated resolutions, where our most advanced customers are solving 80% or 90% of their interactions through an AI agent…the solutions can now really execute the promise of personalized customer service that’s proactive and actually solve a customer’s problem… .”
“If a customer can get their problem or opportunity solved very quickly and efficiently through a bot or an agent, they’re happy to do that,” he further says. “The problem is the technology was a little kludgy in the past. I think we’ve got over that hill now where it can really satisfy companies and consumers.”
AI Scaling Hits Its Limits
Power caps, rising token costs, and inference delays are reshaping enterprise AI. Join our exclusive salon to discover how top teams are:
- Turning energy into a strategic advantage
- Architecting efficient inference for real throughput gains
- Unlocking competitive ROI with sustainable AI systems
Secure your spot to stay ahead: https://bit.ly/4mwGngO
The next-generation AI agents
Zendesk’s new AI agents use Zendesk AI, a solution powered by OpenAI, off-the-shelf proprietary models trained by Zendesk and a set of grounding and control mechanisms. The company claims its bots will autonomously interact with customers, comprehend their needs and deliver the necessary resolution. The “sophisticated” AI-powered bots can be integrated with any knowledge base and customized to respond to “the most intricate use cases.”
Using Zendesk AI’s “dataset and expertise allows us to train CX-specific models for foundational tasks like intent, sentiment and entity detection,” Cristina Fonseca, the company’s head of AI, said in a statement. “This ensures our AI delivers 30 percent better accuracy in understanding customers than generic models. We also designed our AI to help customers embrace [gen AI] step-by-step while building trust in the technology.”
However, these bots are only one part of the company’s announcements today. Zendesk is also launching its Agent copilot, an intelligent guide trained on past customer service experiences to improve workflows, address customer needs and suggest ways to resolve future interactions. It studies the tickets that human agents are working on and proactively suggests answers before the person knows how to respond.
Eggemeier describes it as putting the human agent first and the AI second, flipping the script that other similar copilots might follow. “If you are interacting with the human agent through a chat, messaging, email, web form or any digital medium, our AI copilot actually suggests an answer for the agent before the agent types anything in.” He shares that the recommended responses are generated based on what a business knows about the customer, pulling insights from customer relationship management tools, analyzing their cookies and relying on other sources.
“It’s giving a response, and the agent is then editing it, approving it, changing the tone and checking for accuracy. I think that’s a fundamental shift in a fundamental difference from our competitors,” he claims.
Throughout the interview with VentureBeat Eggemeier emphasized that he doesn’t envision a future in which AI replaces humans in customer service. “We have a view that there is always a human, even if you use your own bot to interact with a company, that there’s a human on the other end. So we need to focus with companies on how we go and deal with that human, whether they’re an employee or consumer, and give them the best experience.”
However, there may be some instances where a bot is best suited. “One of the biggest problems with human agents is that they do a bunch of remedial-like menial tasks over and over again,” Eggemeier said. These activities don’t contribute to job satisfaction. “We think the AI agent will offload a lot of that. And the agent copilot will give you that so you’re doing more creative problem-solving.”
In addition, the agent and copilot will provide customer service agents with the necessary context to strike a more empathetic tone, saving customers time from having to restate their cases repeatedly.
Workforce experience management
For businesses using human and AI agents, Zendesk has a scheduling feature to ensure optimal staffing within the call center. Administrators can use this workforce management tool to determine the number of human agents needed based on factors such as holidays and inclement weather. Supervisors can route calls to the AI agents if there aren’t enough human agents. Eggemeier says it’s replacing the traditional method of scheduling which is done using an Excel spreadsheet.
But this suite of tools goes beyond simple scheduling. Using its predictive algorithm, Zendesk can determine which agent—human or AI—will provide the best resolution and reroute the call accordingly. For example, if you’re calling or emailing to cancel a subscription, Zendesk’s AI could determine that a human agent might better retain you as a customer than an AI agent.
“Now, most of the time you just want to go to an AI agent because it’s most efficient, but if [a business] has extra capacity right now—you know that in real-time—and you’ve got some high-value customers, you’ll start routing them because you have the scheduling to those human agents because you have that capacity,” Eggemeier said.
“Conversely, sometimes you won’t have enough capacity, even with the best planning. And if someone has to wait on a chat for five minutes to get a human agent, even though the human agent deals with that kind of interaction better, they’re better off getting an AI agent to solve it in a minute.”
Ensuring AI delivers results
The last component introduced today is powered by Klaus, a startup Zendesk acquired in January. The customer service platform will now evaluate 100 percent of all AI Agent interactions and use AI to identify those interactions that require human intervention to minimize churn risk, repair incorrect workflows, and provide knowledge center updates.
Eggemeier admits that humans and bots make mistakes, but better answers can be generated when both are linked together, resulting in fewer hallucinations. He boasts that Zendesk’s quality assurance tool will ensure “higher fidelity, fewer mistakes from humans, fewer hallucinations from AI, and we’re going to learn from those very quickly before one gets out of control.”
He reiterated the 100 percent interaction figure, telling VentureBeat that other providers are sampling 1-2% of interactions. By scanning every conversation, Zendesk’s routing engine can assess if AI agents are doing a better job or are humans.
Creating an open CX platform
Like many tech companies, Zendesk doesn’t back one particular AI model or service. It’s receptive to all. In a separate announcement at this year’s Relate conference, Zendesk reveals it’s supporting Amazon Bedrock and Anthropic. The former builds on an existing relationship between the two companies while the latter provides Zendesk’s 100,000 customers greater flexibility to use the large language model (LLM) of their choice to tackle customer interactions.
“AI is transforming customer and employee interactions, and soon, it will shape virtually all of them. This transformation requires mastery and efficient deployment of the best LLM technology available,” Adrian McDermott, Zendesk’s CTO said in a statement. “Our long-standing partnership with AWS and work with Anthropic means our customers have a CX platform and choice of powerful LLMs to help them set a new standard for service, with AI and automation providing support quickly and effortlessly.”
Eggemeier teases that more back-end system integrations will be revealed in the coming months to leverage the Zendesk AI better. However, he declined to provide specifics.
