October 14, 2025, 12:16 pm IDT
The fragmented customer journey, where a business presents “multiple personalities” across different touchpoints, is a relic poised for profound transformation by artificial intelligence. This insight, shared by Cresta CEO Ping Wu, sets the stage for a compelling discussion on the future of customer experience, as he spoke with Sequoia Capital’s Sonya Huang and Doug Leone on the “Training Data” podcast. The interview explored not just the mechanics of contact center AI, but the philosophical shift from scarcity to an abundance mindset, envisioning entirely new paradigms of interaction.
Ping Wu, who previously built Google’s contact center business before leading Cresta, champions a dual approach to AI implementation: automating processes that are readily adaptable while simultaneously augmenting human agents with intelligent assistance. This perspective directly challenges the often-cited fear of wholesale human displacement, instead framing AI as a catalyst for unprecedented personalization and continuous customer conversations. Doug Leone, a seasoned investor, underscored this nuanced view, stating that the precise percentage of human labor replaced is less critical than “the speed of adoption” of AI technologies within the vast, often entrenched, contact center market.
The contact center, a colossal industry employing 17-20 million human agents globally, has historically been ripe for technological disruption, yet has remained surprisingly static. Wu highlighted its inherent inefficiencies: customers endure long wait times, agents face high attrition rates (35-40% annually, sometimes over 100% during the pandemic), and businesses constantly seek to “do more with less.” This universal dissatisfaction, according to Wu, represents a monumental opportunity for AI to introduce an “abundance” of solutions. Large Language Models (LLMs) are identified as the “perfect tool” to weave together disparate interactions—from sales calls to digital chats and emails—into a cohesive, personalized narrative throughout the entire customer lifecycle, a level of seamlessness previously unattainable.
A key insight emerging from the discussion is that the immediate future of contact center AI isn’t about full automation, but strategic augmentation. Wu pointed out that Gartner research suggests not a single Fortune 500 company will achieve a fully human-less contact center within the next five years. This realistic outlook acknowledges the complexity of legacy systems and the deeply ingrained processes that characterize large enterprises. The true revolution, as Leone emphasized, lies not in the final degree of automation, but in the velocity at which these capabilities are integrated and adopted across an organization.
Cresta’s approach epitomizes this blended strategy, offering both autonomous digital agents for routine tasks and AI-powered tools that assist human agents in real-time. The goal is to offload “low-emotion” interactions that neither customer nor business wants to have, such as authentication or basic data entry. Beyond mere efficiency, AI is expected to enable entirely new forms of customer engagement, like direct voice interaction with airline apps or transforming synchronous calls into asynchronous, self-service resolutions. This shift liberates human agents to focus on complex, high-value interactions, improving both employee satisfaction and overall customer experience.
Related Reading
The implementation of advanced Contact Center AI, however, is not without significant technical hurdles. Wu detailed the intricate technical stack required, including streaming end-to-end bidirectional audio, orchestrating over 20 different models (speech-to-text, noise cancellation, turn detection, text generation, guardrails, and more), and achieving near real-time latency (below 800 milliseconds). Furthermore, legacy IT infrastructure, often lacking modern APIs and optimized for human-centric graphical interfaces, presents a substantial integration challenge that requires more than just advanced AI models. These are not solely AI problems but fundamental infrastructure issues that demand attention before full AI potential can be unlocked.
Leone also offered a perspective on where value accrues in the broader AI market. He contends that value will consistently flow “up” to the application layer, rather than solely resting with foundational models or compute infrastructure. This means that companies that effectively leverage AI models to solve specific, real-world business problems and deliver tangible value to customers will be the ultimate winners. The “sizzle” of cutting-edge AI must be backed by “steak”—robust, practical applications that address genuine needs.