Inside Zendesk’s dual AI leap: From reliable agents to real-time intelligence with GPT-5 and HyperArc
Presented by Zendesk
Agentic AI is currently transforming three key areas of work: creative, coding and support, says Shashi Upadhyay, president of engineering, AI and product at Zendesk. But he notes that support poses a clear challenge.
“Support is special because you put an autonomous AI agent right in front of your customer,” says Upadhyay. “You need to be able to trust that it is doing the right thing for the customer and by the customer. Every step forward in AI should make the service more reliable for both customers and human agents.”
Zendesk, recently named a leader in the Gartner Magic Quadrant 2025 for the CRM Customer Engagement Center, started deploying AI agents about a year and a half ago. Since then, they’ve seen AI agents resolve nearly 80% of all incoming customer requests on their own. For the remaining 20%, the AI agent can hand it over to a human to solve the more complex problems.
“Autonomous AI agents work 24/7, with no wait times. You have a problem; they give an answer right away. It all adds up,” he says. “Not only do you get higher resolutions and higher automation, but you can also improve the CSAT at the same time. Because 80% is such a promising number and the results are so solid, we think it’s only a matter of time before everyone adopts this technology. We’re already seeing that across the board.”
The company’s efforts to improve usability, depth of insight and time-to-value for organizations of all sizes require continued testing, the integration of advanced models like ChatGPT-5 and a major upgrade in analytics capabilities and real-time, AI-powered insights with the acquisition of HyperArc, an AI-native analytics platform.
Design, test and implement a better agent
“Especially in an enabling context, it is important that AI agents behave consistently in accordance with the company brand, policies and regulatory requirements you may have,” says Upadhyay. “We test every agent, every model all the time with all our customers. We do it before we release it and we do it after we release it, in five categories.”
These categories – automation speed, execution, precision, latency and security – form the basis of Zendesk’s ongoing benchmarking program. Each model is scored based on how accurately it solves problems, how well it follows instructions, how quickly it responds, and whether it stays within clearly defined guardrails. The goal is not only to make AI faster, but also to make it reliable, responsible, and aligned with the standards that define great customer service.
That testing is enhanced by Zendesk’s QA agent: an automated monitor that continuously monitors every conversation. If an exchange starts to veer off course, both in tone and accuracy, the system immediately signals this and alerts a human agent to intervene. It’s an extra layer of assurance that keeps the customer experience on track, even when AI is the first line of support.
GPT-5 for next level agents
In the world of support and service, the move from simple chatbots that answer basic questions or solve straightforward problems to agents that actually take action is game-changing. An agent who can understand that a customer wants to return an item, confirm its eligibility for return, process the return, and issue a refund is a powerful upgrade. With the introduction of ChatGPT-5, Zendesk saw an opportunity to integrate that capability into its Resolution Platform.
“We have worked works closely with OpenAI because GPT-5 was a pretty big improvement in the model’s capabilities, ranging from being able to answer questions to being able to reason and take action,” says Upadhyay. “First, it is much better at solving problems on its own. Secondly, it is much better at understanding your intent, which improves the customer experience by making you feel understood. Last but not least, it has a reliability of more than 95% in terms of correct execution.”
These gains span Zendesk’s AI agents, Copilot, and App Builder. GPT-5 reduces workflow errors by 30%, thanks to its ability to adapt to unexpected complexity without losing context, and reduces fallback escalations by more than 20%, with more complete and accurate responses. The result: faster resolutions, fewer transfers, and AI that behaves more like a seasoned support professional than a script assistant.
Additionally, GPT-5 is better at handling ambiguity and can clarify vague customer input, improving routing and increasing automated workflows in more than 65% of calls. It is more accurate in five languages and makes agents more productive with more concise, contextually relevant responses that align with tone guidelines.
And in App Builder, GPT-5 delivered 25% to 30% faster overall performance, with more rapid iterations per minute, accelerating development workflows for app builders.
Filling the analytics gap
Traditionally, support analytics has focused on structured data, the kind that fits neatly into a table: when a ticket was opened, who handled it, how long it took to resolve it, and when it was closed. But the most valuable insights are often in unstructured data: the conversations themselves, spread across email, chat, voice and messaging apps like WhatsApp.
“Customers often don’t realize how much intelligence goes into their support interactions,” says Upadhyay. “What we’re aiming for with analytics are ways we can improve the entire business with the insights contained in the supporting data.”
To surface these deeper insights, Zendesk turned to HyperArc, an AI-native analytics company known for its proprietary HyperGraph engine and generative AI-powered insights. The acquisition revitalized Explore, Zendesk’s analytics platform, and transformed it into a modern solution capable of joining structured and unstructured data, supporting conversational interfaces, and leveraging persistent memory to use past interactions as context for new queries.
“Your support interactions tell you everything that’s currently not working in your business. All that information is in the millions of tickets you’ve collected over time,” says Upadhyay. “We wanted to make that fully visible. Now we have this genius AI agent that can analyze it all and come up with explicit recommendations. That not only improves support, but it improves the entire business.”
That visibility now translates into useful information. The system can identify where problems are most persistent, identify the patterns behind them and suggest ways to solve them. It can even anticipate problems before they happen. For example, during high-pressure events such as Black Friday, it can analyze historical data to identify recurring issues, predict where new bottlenecks may occur and recommend preventative measures, turning reactive support into a proactive strategy.
“That’s where HyperArc excels,” says Upadhyay. Not only does it help you understand the past, it also helps you plan better for the future.”
By integrating HyperArc’s AI-native intelligence, Zendesk moves customer service toward continuous learning – where every interaction builds trust and improves performance, paving the way for AI that can see what’s coming next.
Sponsored articles are content produced by a company that pays for the post or has a business relationship with VentureBeat, and is always clearly marked. For more information please contact sales@venturebeat.com.
