Sierra Co-Founder on AI Agents, Outcome Pricing, and ROI

AI agents transition from proof-of-concept to production

AI agents designed to complete tasks rather than simply answer questions are beginning to move from demonstration environments into actual business operations, according to Sierra co-founder Clay Bavor. In a conversation with CNBC's Arjun Kharpal, Bavor outlined how his company is deploying customer-facing AI agents in customer service, sales, and support functions.

Sierra builds and tests these agents before they interact with real customers, a process Bavor described as critical to ensuring reliability in production environments. The company's approach reflects broader industry concerns about the gap between AI capabilities in controlled settings and their performance when deployed at scale.

Why it matters

As enterprises move beyond chatbot experiments, they're demanding clearer metrics for AI return on investment. This pressure is forcing AI vendors to reconsider traditional software pricing models, potentially shifting from subscription fees to outcome-based arrangements that tie costs directly to business results. That change could fundamentally alter revenue models across the software industry.

The pricing model challenge

Bavor addressed growing demand from companies for better ways to measure AI's return on investment. This need for accountability is driving interest in outcome-based pricing models, where customers pay based on results rather than seats or usage tiers. Such a shift would represent a significant departure from conventional SaaS pricing structures that have dominated enterprise software for years.

The conversation touched on rising AI token costs, which continue to be a factor in deployment economics, as well as coding agents that assist with software development tasks.

The last-mile problem

Bavor identified enterprise AI deployment's "last mile" as particularly challenging. This final stage—moving from working prototypes to systems that operate reliably in complex business environments—remains a significant hurdle for organizations implementing AI agents. The gap between demo and deployment involves integration with existing systems, handling edge cases, and maintaining consistent performance under real-world conditions.

The discussion, part of CNBC's "The Tech Download" series, comes as businesses increasingly evaluate AI agents for tasks that require multi-step reasoning and action, not just information retrieval. Customer service represents one of the most active areas for this technology, where agents can potentially handle inquiries, process requests, and escalate issues without human intervention.

These details were first reported by CNBC in their interview with Bavor.