Why AI Will Transform Customer Experience: Cresta CEO Ping Wu and Sequoia’s Doug Leone

Market Structure & Opportunity

• The contact center market encompasses 17–20 million human agents globally, with a software market valued in the tens of billions.
• Only 25% of contact center interactions are purely support-focused; the remaining 75% are revenue-generating activities (selling, collections, retention).
• Historically, the industry suffers from massive attrition rates of 35–40%, with some periods exceeding 100% turnover due to high stress.
• Customer experience is currently fragmented, with sales, onboarding, and service departments operating as disconnected "personalities."

Automation Predictions & Timelines

• No Fortune 500 company is expected to operate a 100% humanless contact center within the next five years.
• Total automation of labor is projected to take 25–30 years due to legacy system constraints (e.g., IBM mainframes, COBOL).
• Full automation of interactions (customer calls) may accelerate faster than labor displacement, potentially within 2–3 years for specific use cases.
• AI agents are expected to surpass human agents in knowledge, language capability, and emotional consistency within 2–3 years.

Cresta's Strategic Approach

• Cresta employs a "meet the customer where they are" strategy, utilizing a hybrid model of human agent assist and autonomous digital agents rather than waiting for full automation.
• The company focuses on "abundance mindset" value creation, such as enabling asynchronous interactions (e.g., AI resolving issues and calling back) that were previously impossible due to staffing limits.
• Value accrual is prioritized at the application layer ("the last mile"), where AI integrates with complex, legacy backend systems that lack APIs.
• The product strategy combines "steak" (operational agent assist and backend integration) with "sizzle" (consumer-facing autonomous agents).

Technology Stack & Latency

• The Voice AI Agent stack streams end-to-end audio bi-directionally, orchestrating over 20 models simultaneously.
• Core models include speech-to-text, noise cancellation, interruption detection, foundational LLMs, and text-to-speech (TTS).
• Guardrails run in parallel to enforce specific business rules (e.g., prohibiting tax advice) and prevent hallucinations.
• The system maintains a latency of under 800 milliseconds to simulate natural human conversation.
• Training utilizes real-world human conversation data to create simulation blueprints and test edge cases before deployment.

Investment Outlook (Doug Leone)

• Current AI investment is compared to the internet/mobile cycles, characterized by rapid shifts from infrastructure to application layer value.
• Sequoia's thesis is that long-term value will accrue near the customer and business user (application layer) rather than at the compute or model layer.
• The AI wave is described as "Industrial Revolution 2.0," representing a non-linear transformation of human workflows rather than simple connectivity.
• Investment discipline involves forcing founders to remove obstacles ("rocks") to achieve growth rates that are linear only after significant scaling, rather than accepting conservative ramp forecasts.

Future Vision

• AI is expected to disappear into workflows, making the distinction between AI and human agents indistinguishable within 20–30 years.
• The primary long-term goal is to create a continuous, personalized customer conversation that spans the entire customer journey, eliminating disjointed departmental experiences.
• The most significant hidden value lies in solving "low emotion value" interactions (e.g., authentication, simple queries) to reduce call volume before they reach human agents.