Bret Taylor Sees AI Agents Repeating the Dotcom Boom

Bret Taylor on AI Agents, Outcome Pricing, and a Dotcom-Style Boom

At a live Decoder event, Sierra CEO and OpenAI chair Bret Taylor explained why he left Salesforce after the ChatGPT moment and dove into building AI agents for customer experiences. Sierra’s platform powers voice and chat agents for brands like ADT and SiriusXM, automating complex flows such as warranty claims and even end-to-end mortgage refinancing without a human in the loop.

Taylor argues the technology shift is as fundamental as the early internet: it reshuffles which vendors win, and creates new markets. That’s why Sierra targets large enterprises—companies with tens of millions of users where reducing the cost of a phone call by orders of magnitude changes economics and customer lifetime value.

A key product and business differentiation: Sierra charges for outcomes. If an AI agent autonomously resolves a case, the customer pays; transfers to humans are free. Taylor sees outcome-based pricing as natural for agents — it aligns incentives and forces vendors to own success, not just sell software.

Beyond simple retrieval Q&A, Sierra emphasizes deterministic and AI-based guardrails for regulated workflows: multilingual transcription accuracy, supervisor models, and domain-specific checks that prevent costly errors. These are the engineering details many teams underestimate, and Taylor says they explain why enterprise rollouts are still hard.

Voice matters. Taylor believes voice will outpace chat because it’s low-friction, accessible, and maps to existing phone-centric industries like telecom and healthcare. He points out that agents turning analog phone systems into digital conversational channels is a paradigm shift as big as putting the internet on the phone.

On models and tech: Sierra fine-tunes rather than pretrains models, orchestrating multiple inference calls and swapping providers to balance cost, latency, and quality. Taylor compares models to databases — practitioners will choose the right model for the job rather than invent foundational models themselves.

Taylor acknowledges the hype: he thinks we’re in an AI bubble that mirrors the dotcom era — massive opportunity alongside many failures. His advice is practical: invest in applied AI companies that deliver outcomes, not performative pilots, and expect an ecosystem of vertical agent vendors to emerge.

Practical implications for CX and tech leaders:

• Prioritize measurable outcomes — tie vendor fees to resolved cases and ROI.
• Design deterministic guardrails and multilingual voice pipelines before wide rollout.
• Treat models as interchangeable infrastructure components and optimize for task-specific tradeoffs.

Taylor’s view is clear: we’ve only begun to see how far AI agents can change customer interactions and software creation. For enterprises, the immediate wins are in reducing contact-center cost, increasing conversation volume, and converting service into competitive advantage. The hard — but valuable — work is integration, monitoring, and compliance.

For leaders asking whether to build or buy, Taylor’s pragmatic answer is to buy applied solutions that deliver outcomes while building internal expertise in agent orchestration. Expect a future where every vertical has specialized agent vendors, and where paying for resolved outcomes is the new normal.