Clay Bavor is the cofounder of Sierra, the enterprise AI company he started with Bret Taylor to build customer-facing agents that solve problems, complete transactions, and reflect a company’s identity. Previously, he spent 18 years at Google, leading Google Workspace’s product and design teams before starting and leading Google Lens, Project Starline, and Google’s augmented- and virtual-reality efforts; he ultimately led Google Labs.
His work on Daydream and immersive computing reflected a lasting interest in interfaces that reduce the distance between human intentions and machine operations. In a 2017 essay, he traced computing’s progression toward conversational systems and experiences embedded in the physical world.
Bavor and Taylor introduced Sierra in 2024, proposing agents that securely connect to business systems, handle requests such as returns and subscription changes, and preserve a company’s brand. Early customers included WeightWatchers, SiriusXM, Sonos, and OluKai.
- The AI architect: Bavor sees customer-facing agents creating a role comparable to the early webmaster: someone who understands the technology, shapes the agent’s identity, and ties its performance to business outcomes. He considers customer-experience professionals particularly well positioned because they understand customer needs and can continuously review and improve interactions. His AI Engineer conversation with Alessio Fanelli also highlights model upgrades, voice interruptions, speaker identification, and the operational complexity behind dependable agents.
- Agent Development Life Cycle: Bavor and Taylor outlined an engineering framework for probabilistic software incorporating simulated conversations, regression testing, human feedback, immutable releases, and deterministic guardrails where policy or compliance requires predictable behavior.
- Voice Personas: Sierra’s 2026 voice-persona launch separates what an agent can do from how it speaks, allowing personality, language, pacing, and conversational style to change while business policies remain consistent.
- Long-horizon agents: With Horizon, Bavor and Taylor extended agent interactions across days or weeks, using persistent context to pursue outcomes such as healthcare scheduling, loan origination, and subscription upgrades.
Bavor expects AI interfaces to move beyond chat into combinations of voice, imagery, video, and wearable devices. His interest in glasses and ambient assistants draws directly on his earlier work in visual and immersive computing.
