Presented by Zendesk
Agentic AI is currently transforming three key areas of work – creativity, coding and support, says Shashi Upadhyay, president of engineering, AI and product at Zendesk. However, he notes that support presents a separate challenge.
“The support is unique because you are giving the customer an autonomous AI agent,” Upadhyay says. “You need to make sure it’s done right for the customer and by the customer. Every step forward in AI should make services more reliable for both customers and employees.”
Zendesk, recently named a Leader in… Gartner Magic Quadrant 2025 for CRM Customer Engagement Center, the implementation of AI agents began about a year and a half ago. They have since seen that AI agents are able to resolve almost 80% of all incoming customer requests on their own. The remaining 20% can be handed over to a human to facilitate solve more elaborate problems.
“Autonomous AI agents work 24/7, without waiting or queuing. You have a problem, they respond immediately. It all adds up,” he says. “Not only do you get higher resolution and more automation, but you can also improve CSAT at the same time. Because 80% is such a promising number and the results are so solid, we believe it’s only a matter of time before everyone adopts this technology. We’re already seeing it around the world.”
The company’s efforts to raise the standard of usability, depth of insight and time to value for organizations of all sizes require continuous testing, integration of advanced models such as ChatGPT-5, and significant modernization of its real-time, AI-powered analytical capabilities and insights through the acquisition of HyperArc, an AI-native analytics platform.
Design, test and deploy a better agent
“Especially in the context of support, it’s important that AI agents behave consistently with the company brand, policies and regulatory requirements you may have,” Upadhyay says. “We test every agent and every model continuously across all of our customers. We do this before we release the product and we do it after the product is released, across five categories.”
These categories – degree of automation, execution, precision, latency, and security – form the basis of Zendesk’s ongoing benchmarking program. Each model is rated based on how accurately it solves problems, how well it follows instructions, how quickly it responds, and whether it stays within clearly defined limits. The goal is not just to accelerate AI – but to make it reliable, accountable and compliant with the standards that define great customer service.
This testing is supported by a Zendesk QA agent, an automated monitor that constantly watches every call. If an exchange begins to veer off course, either in tone or accuracy, the system immediately flags it and alerts the human agent to step in. This is an additional level of assurance that keeps customer service on track, even when the first line of support is artificial intelligence.
GPT-5 for higher-level agents
In the world of support and services, the shift from plain chatbots that answer basic queries or solve plain problems to agents that actually take action is groundbreaking. An agent who can understand that a customer wants to return a product, confirm whether it qualifies for a return, process the return, and issue a refund is a powerful improvement. With the introduction of ChatGPT-5, Zendesk saw an opportunity to integrate this capability into its troubleshooting platform.
“We worked very closely with OpenAI because GPT-5 has enabled significant improvements in the model’s capabilities, from the ability to answer questions to the ability to reason and take action,” says Upadhyay. “First, it is much better at autonomous problem solving. Second, we understand your intentions much better, which improves the customer experience because you feel understood. And last but not least, the operational reliability is over 95%.
These benefits translate to Zendesk’s AI agents, Copilot and App Builder. GPT-5 reduces workflow failures by 30% by adapting to unexpected complexity without losing context, and reduces emergency escalation by over 20% by providing more complete and accurate responses. The result: faster resolutions, fewer handoffs, and AI that behaves more like a seasoned help desk specialist than a script assistant.
Additionally, GPT-5 better handles ambiguity and is able to clarify unclear customer input, which improves routing and increases workflow automation in over 65% of calls. Delivers greater accuracy in five languages and increases agent productivity with more concise, contextually appropriate responses that adhere to tone guidelines.
And in App Builder, GPT-5 delivers 25% to 30% higher overall performance, with more fast iterations per minute, accelerating your app development workflow.
Filling the analytical 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 information often resides in unstructured data—in the conversations themselves, distributed across email, chat, voice apps, and messaging apps like WhatsApp.
“Customers often don’t realize how much intelligence there is in their interactions with support,” says Upadhyay. “When it comes to analytics, we focus on how we can improve the entire business with the insights from support data.”
To mine these deeper insights, Zendesk turned to HyperArc, an AI analytics company known for its proprietary HyperGraph engine and generative AI-powered insights. The acquisition revitalized Explore, Zendesk’s analytics platform, into a current solution capable of combining structured and unstructured data, supporting conversational interfaces, and using persistent storage to leverage past interactions as context for modern queries.
“Through your support interactions, you’ll learn everything that’s currently not working in your business. All of this information is embedded in the millions of tickets you’ve collected over time,” says Upadhyay. “We wanted it to be completely visible. Now we have a brilliant AI agent who can analyze it all and come back with clear recommendations. It doesn’t just improve support. It improves the entire company.”
This visibility now translates into actionable intelligence. The system can pinpoint where problems are most persistent, identify patterns behind them, and suggest ways to solve them. He can even anticipate problems before they occur. For example, during high-pressure events like Black Friday, it can analyze historical data to highlight recurring issues, predict where modern bottlenecks may arise, and recommend preventive measures – turning reactive support into a proactive strategy.
“This is where HyperArc shines,” says Upadhyay. It not only helps you understand the past, but helps you plan better for the future.
By integrating HyperArc AI-native intelligence, Zendesk moves customer service towards continuous learning, where each interaction builds trust and improves performance, setting the stage for AI to see what’s next.
Sponsored articles are content created by a company that pays to publish or has a business relationship with VentureBeat and is always clearly marked. For more information, please contact us sales@venturebeat.com.