In the era of instant communication, customer service is no longer a postscript to business—it is the business. Yet, even as enterprises across the Asia Pacific experiment with AI pilots, few have managed to scale from chatbots that merely converse to intelligent agents that truly resolve issues. For many, the challenge lies not in ambition but in architecture, governance, and the ability to balance technological sophistication with human empathy.
Mitch Young, Senior Vice President, Asia Pacific at Zendesk, has a front-row seat to this transformation. Leading one of the most dynamic regions in the world, he believes the future of AI-powered customer experience (CX) will be defined by resolution-first intelligence—systems that integrate seamlessly, learn continuously, and deliver trust as much as speed. In this conversation, he breaks down why AI success in CX isn’t just about adopting new tools, but about aligning people, processes, and platforms. From the emergence of Agentic AI to outcome-based pricing and whitebox transparency, Young offers an inside look at how Zendesk is helping enterprises move beyond pilots to scalable, ethical, and measurable AI transformation.
Senior Vice President – Asia Pacific
Zendesk
CIO&Leader: Most organisations experiment with AI pilots, but few scale successfully. From a technology standpoint, what are the most significant roadblocks APAC enterprises face when scaling AI-powered customer experience (CX) systems?
Mitch Young: In conversations with leaders across APAC, I’m seeing a strong appetite for scaling AI, where real progress stems not just from having the right technological foundations–such as modern, trusted data management, simplified integrations, and agile infrastructure–but also from investing in people and culture. Those effectively scaling their AI are nurturing talent, addressing skill gaps, and choosing AI solutions that integrate seamlessly into existing workflows to deliver immediate value without lengthy deployments or complex setups. Additionally, the introduction of outcome-based pricing, where customers only pay for issues successfully resolved by AI, offers organizations greater transparency and alignment by directly linking investments to measurable business outcomes.
From my perspective, customers who have considered building AI for service in-house, rather than adopting a proven, purpose-built solution, often underestimate the complexities — for example, the heavy engineering overhead required to meet compliance standards, among other reasons. AI for service isn’t just another machine learning project. Choosing a platform that has been shaped by billions of real-world service interactions and brings AI agents pre-integrated into a unified resolution environment can make a significant difference. In most cases, implementing a ready-made solution enables teams to accelerate value, leverage best practices, and stay focused on their core business and what makes them unique, leaving the technical complexity to expert partners.
By addressing not only these technical and strategic choices upfront, but also the domain expertise behind a solution, organizations are far better positioned to move beyond pilots and unlock the sustainable benefits of AI at scale.
CIO&Leader: How does Agentic AI differ from traditional AI chatbots in terms of architecture and capability, particularly when handling complex, multi-intent customer interactions
Mitch Young: Traditional chatbots can handle basic, single-intent queries but often reach their limits as soon as a customer asks something more nuanced, changes intent mid-conversation, or expects help across multiple channels. We’ve all experienced the frustration of being kicked out of a rigid flow and forced to start over when our needs fell outside a bot’s predefined script–something all too common in early-generation chatbots.
Agentic AI, by contrast, operates as a well-orchestrated network of intelligent agents engineered to coordinate actions, maintain context, and deliver resolution across even the most complex, multi-intent journeys. What sets Zendesk’s Agentic AI apart is its unwavering focus on resolution. No matter how a conversation shifts or evolves, our AI agent stays engaged until the customer’s reason for reaching out is fully resolved. That means orchestrating actions across channels and collaborating with human agents in alignment with defined business processes and objectives, even as interactions shift from messaging to voice or email. Crucially, agentic AI adapts with experience — learning from each interaction and refining its performance, always within clear boundaries of governance and transparency.
Advancements in foundational models, such as OpenAI’s GPT-5, have raised the bar for managing ambiguity and context. Zendesk’s integration of GPT-5 has seen significantly improved reliability in handling ambiguity and shifting context, resulting in a reduction of over 20% in human escalations. This evolution brings customers the resolution, speed, and empathy they expect, regardless of complexity or touchpoint.
CIO&Leader: As enterprises modernize their CX infrastructure, how should they integrate AI into existing legacy systems without disrupting service continuity or data flow?
Mitch Young: When modernizing CX infrastructure, there are several viable approaches, from end-to-end transformation to layering AI onto existing systems. The choice depends on a host of factors, including business priorities, resource constraints, and future growth plans.
For those integrating with legacy environments, it’s helpful to start by assessing infrastructure compatibility and understanding the underlying data architecture. This can help pinpoint where, and how, AI can create the most value, while also reducing risk during integration. Many organizations see early gains by applying AI to high-volume, repetitive service areas that don’t require major re-engineering. Solutions that support seamless integration with existing knowledge systems, low- or no-code workflows, and robust, built-in quality assurance can further simplify the process.
A unified platform that brings together ticketing, AI, workforce management, and quality assurance reduces integration overhead and complexity. With built-in QA monitoring of 100% of AI and human interactions across all channels, organizations can identify issues before they disrupt workflows or impact customers, allowing teams to address challenges proactively.
Whatever the path, a purposeful, outcome-oriented approach helps future-proof CX without unnecessary disruption.
CIO&Leader: AI thrives on data quality and contextual learning. What technical approaches or frameworks does Zendesk recommend to ensure continuous model improvement while maintaining privacy and compliance?
Mitch Young: We take a whitebox approach to AI at Zendesk: users can see how every AI decision is made, with access to reasoning steps, prompt testing, and fallback logic for ongoing auditing and fine-tuning.
AI’s growing role in customer service presents an opportunity to set higher standards for transparency, reliability, and real-time adaptability, given the significant importance of trust and resolution to customers. That’s why our models are rigorously tested against real-world scenarios using golden datasets to ensure reliability, accuracy, and compliance. Maintaining a robust knowledge foundation keeps the AI grounded in trusted sources, so responses remain accurate and contextually relevant.
A platform-based approach enables the embedding of operational guardrails — such as intent-layer pre-routing, structured logging, and clearly defined escalation paths — across every channel, workflow, and brand. This brings greater control, so out-of-policy responses are avoided, and complex cases can be seamlessly routed to human agents for resolution.
It’s essential to treat AI models as nondeterministic tools operating within a controlled, auditable system, never as standalone decision-makers. With continuous monitoring across offline and live metrics, organizations can build trust, uphold compliance, and adapt confidently as AI capabilities evolve.
CIO&Leader: Beyond operational efficiency, what metrics or KPIs should CIOs and CISOs track to evaluate the tangible ROI of large-scale AI deployment in customer service?
Mitch Young: Across APAC and in India, customer expectations are being reshaped by the advent of AI and a culture that increasingly values instant gratification. Speed and operational efficiency are still critical–especially in our fast-moving APAC markets–but they only show part of the picture. Today’s customers expect not only quick responses but also accurate and meaningful resolutions that reinforce their trust in a brand. Our research indicates that as many as 70% of Indian consumers are likely to switch brands after just one negative experience, and 85% now view customer service not as just a touchpoint, but as a battleground for their loyalty.
When evaluating the tangible ROI of AI, outcome-driven KPIs provide a more meaningful picture of business impact than simply counting interactions. It’s essential to look beyond traditional metrics, such as Customer Satisfaction, Net Promoter Score, and Average Handle Time, and be mindful of potential hidden costs as you scale AI. Some usage-based pricing models — where you’re charged for every interaction or session — can rapidly inflate costs and mask inefficiencies if unresolved or incomplete cases are counted as successful.
In contrast, a resolution-based approach ties investment directly to the effective resolution of customer issues and incentivizes vendors to make their AI as effective as possible, since our success is directly linked to customer outcomes.
At Zendesk, we place particular value on the quality and consistency of resolutions–because, ultimately, that’s what customers value most. Aligning both measurement and commercial models to actual outcomes helps drive loyalty, trust, and long-term differentiation as AI becomes more embedded in service environments.
CIO&Leader: With APAC countries moving toward stricter AI governance, how is Zendesk ensuring transparency, auditability, and ethical alignment in its AI-driven Resolution Platform?
Mitch Young: Ensuring responsible, transparent AI is foundational to how we develop and deploy AI. We achieve this through a comprehensive governance program that centralizes risk assessment, compliance controls, and thorough model evaluation to ensure responsible data handling and behavior.
Oversight is woven into our operations, with a dedicated AI Risk Workgroup that actively evaluates emerging risks and models, and a cross-functional AI Governance Executive Committee, drawing expertise from legal, security, product, engineering, marketing, and customer success, guiding overall policy and alignment with global standards.
Zendesk’s responsible AI framework embeds transparency and safety at every stage of development and deployment: from prompt shielding and grounding responses with trusted knowledge sources to rigorous stress-testing and moderation, building on our whitebox AI principles mentioned earlier. This means every AI-driven resolution can be traced, reviewed, and refined.
We have also achieved ISO 42001 certification — the first CX company to do so — providing independent assurance that our people, processes, and controls meet international standards. Customers also retain operational visibility and control, allowing them to deliver AI-driven services with confidence, knowing that every experience is handled transparently and responsibly.