Cheeky Pint Podcast: Bret Taylor of Sierra on AI Agents, Outcome-Based Pricing, and the OpenAI Board

Host: Patrick Collison (of Stripe)
Guest: Bret Taylor (CEO of Sierra, Chairman of the OpenAI Board)
Date: March 10, 2026

Episode Overview

This lively, in-depth conversation between Patrick Collison and Bret Taylor covers the real-world rise of AI agents, practical challenges of deploying AI in large organizations, the shifting ground of enterprise software, and the rapid evolution of both AI business models and engineering practices. Patrick and Bret—both deeply embedded in Silicon Valley—discuss how companies like Sierra are changing customer service, the broader implications of agentic computing, the paradoxes of software’s future, and the organizational impacts of AI, from codebases to boardrooms.

Key Discussion Points and Insights

1. The State of Consumer AI Agents

[00:30-05:45]

• OpenClaw and the Rise of Open Source AI Agents

  • OpenClaw, a “semi-rogue open-source” agent with janky memory (Markdown files), represents a chaotic but compelling use of AI in 2026.
  • Bret Taylor: “The whole thing is fascinating… Instead you’re chatting over WhatsApp with a thing on a Mac Mini that is mildly unhinged and insecure.” [00:33]

• Memory and Context

  • Industry AI products like ChatGPT and Gemini still lack native memory between sessions, while OpenClaw’s primitive memory system is “like the movie Memento.”
  • The counterintuitive value of messy, codebase-like memory structures for AI agents: “Maybe mimicking a code base is actually the best way to make a general-purpose agent work.” [04:16, Bret Taylor]

2. Engineering with and for AI Agents

[05:31-10:00]

• The Terminal Renaissance:

  • Engineers’ long-standing preference for UNIX-like tools and file structures translates surprisingly well to AI agent harnesses.
  • Documentation is resurging as a necessary byproduct of agent workflows (“It would be the greatest irony if software engineering agents made all of us just write documentation…” [06:50, Bret Taylor]).

• Letting Go of Code:

  • Emotional shift for engineers moving from hand-crafted code to outcome-oriented engineering with AI:
    • “I have a hard time not caring... But I've been trying to force myself to not care because I feel like I won't be a self-actualized software engineer in the future if I'm too precious about that artifact…” [07:26, Bret Taylor]
  • Ongoing search for new tools and UIs for future programming, as code becomes less central.

3. Single-Agent vs. Multi-Agent Systems

[09:57-11:55]

• Skepticism about Multi-Agent Protocols (like MCP):
  • Most multi-agent architectures look great on a whiteboard but fail in practice due to lack of context-sharing and the resulting robotic outputs.
  • Messier, more context-rich approaches (à la OpenClaw with markdown context directories) feel more “human” and functional right now.

4. “What’s Old is New”: Return to the Command Line and Harnesses

[11:55-16:23]

• API & Harness Futures:
  • Classic software metaphors—SSH, tail, grep—are reemerging as highly ergonomic for agent-driven interfaces.
  • Future applications may offer both a human-facing web app and an "agent harness"—an interface expressly built for agents to access skills, documentation, and take actions beyond traditional REST APIs.

5. Sierra: Building Real AI Agents for Customer Service

[20:14-23:47]

• What is Sierra?

  • AI agents for customer experience (call centers, chat, digital interfaces).
  • Powers major brands like Cigna, SiriusXM, Sofi, Rocket Mortgage.
  • Extraordinary growth: $100M ARR in 7 quarters; $165M currently.

• How Companies Adopt AI Agents

  • Start with limited use cases, typically by channel (phone/chat), then expand as confidence and results grow.
  • Unification of call center and digital teams as even “analog” channels become digital.

• Future Vision:

  • AI agents as the primary, branded interface for most customer interactions—surpassing both website and app, even including phone-based digital service.

6. AI Agents’ Impact on Customer Service Economics

[24:27-36:26]

• Cost Transformation:

  • Massive reduction in per-case costs (from ~$10-20 per interaction to pennies).
  • New opportunities to offer high-quality support to all customers, not just high-value ones.

• Beyond Cost: Lifting Customer Experience

  • Companies see customer satisfaction rise as agents automate routine queries and free human agents for tougher, more rewarding cases.
  • Outcome: volume of customer conversations increases, not just efficiency; offers companies a competitive imperative rather than just a cost advantage.

7. AI as the New UI for Commerce and Beyond

[27:11-31:22]

• Websites and Forms as Historical Artifacts:

  • "Maybe navigating websites and filling out forms was like a bit of a moment in time." [27:52, Patrick Collison]
  • Conversational interfaces and agents expand the accessible “front door,” subsuming (not eliminating) legacy UI modalities.

• Redefining Digital Market Share:

  • Digital devices accumulate (PC, smartphones, speakers), but AI agents are now omnichannel—interacting over phone, chat, websites, even smart speakers.
  • Potential for less “addictive” tech if screen-based interaction declines.

8. From Usage-Based to Outcome-Based Pricing

[61:35-68:35]

• Sierra’s Business Model:

  • Outcome-based pricing: e.g., only charging if an AI agent fully resolves a case, rather than per usage/token.
  • Aligns incentives: “If an agent's outcome is measurable, it's a really compelling way to both for clients obviously because it's aligned with their business. But it’s also quite disruptive…” [61:35, Bret Taylor]
  • Moves the software industry towards more accountability and away from “implementation blame games.”

• Practical Constraints:

  • Not all agent tasks are outcome-measurable yet; fallback to usage-based in complex product cases.

9. Enterprise AI Adoption, Productivity, and Organizational Change

[69:13-87:35]

• Applied AI’s Enduring Value:

  • Despite rapid progress in foundational models, huge swathes of “boring but important” business processes remain to be agentified—“if we paused model development, we'd still have trillions of dollars of economic value… yet to be realized.” [71:54, Bret Taylor]
  • Most companies want solutions, not raw models.

• Uncertainty in SaaS Market Value:

  • Investors sense increased risk (“more uncertainty now than there’s ever been” [50:50, Bret Taylor]) due to the potential obsolescence of traditional code- and integration-heavy SaaS companies.

• Process vs. Person as the AI Productivity Unit:

  • AI is better at automating processes than substituting for whole jobs, especially as most organizations aren’t yet restructured for process-centric optimization.
  • “The atomic unit of productivity in AI is a process, not a person.” [75:11, Bret Taylor]

• Generalists Ascendant:

  • High-agency, customer-caring, multi-disciplinary employees are empowered by AI, upending the drift toward overspecialization.
  • “Those people are truly worth a thousand x other people. ... In a world of AI agents, those generalists ... can end up more powerful in the Silicon Valley company.” [85:01 & 87:35, Bret Taylor]
  • "I've really noticed that at Stripe." [87:35, Patrick Collison]

10. Twitter, OpenAI Boards, and Leading Through Inflection Points

[90:13-98:16]

• Reflections on the Twitter Board and Elon’s Takeover:

  • “I realized I didn’t love that very much… I’m like a builder.” [90:32, Bret Taylor]
  • Agency and careful team structure matter more than just headcount in software outcomes.

• OpenAI’s Nonprofit Board:

  • Unique experience serving with a fiduciary "duty to a mission": “Your sole duty is to ensure that artificial general intelligence benefits humanity. That’s really different.” [96:46, Bret Taylor]
  • Handpicked diverse board post-crisis, blending safety, economics, and technical depth.

Notable Quotes & Memorable Moments

• Bret Taylor on AI Agents:

  • “Maybe mimicking a code base is actually the best way to make a general-purpose agent work.” [04:16]
  • “It would be the greatest irony if software engineering agents made all of us just write documentation the whole time…” [06:50]
  • “I'm trying to force myself to not be emotionally attached to the code, which is very hard for me...” [09:46]
  • “I think the atomic unit of productivity in AI is a process, not a person.” [75:11]
  • “In a world of AI agents, those generalists ... can end up more powerful in the Silicon Valley company.” [85:01]
  • “If we paused model development, we'd still have trillions of dollars of economic value… yet to be realized.” [71:54]
  • On being on high-drama boards: “Have you considered that you are the problem? You are bringing the drama.” [95:15, Patrick Collison; laughter ensues]

• Patrick Collison on Tech Shifts:

  • “Maybe navigating websites and filling out forms was like a bit of a moment in time.” [27:52]
  • "I've really noticed that at Stripe..." (about generalists being empowered by AI) [87:35]

Important Timestamps

| Timestamp | Segment | |---------------|-----------------------------------------------------| | 00:30-05:45 | The janky state of consumer AI agents and memory | | 16:23-20:14 | Real-world agent vs. protocol (healthcare example) | | 20:14-24:17 | Sierra's business, customers, and ARR growth | | 24:27-36:26 | Impact of AI agents on service cost & experience | | 61:35-68:35 | Outcome-based pricing & business models in AI | | 75:11-83:10 | Applied AI, process-centric productivity | | 87:35-90:13 | Empowerment of high-agency generalists | | 90:13-95:15 | Twitter Board, team size, and organizational design | | 96:45-98:16 | OpenAI board, fiduciary duty to mission | | 98:20-101:18 | AI in 2026: Key predictions |

AI, Agents, and the Enterprise: The 2026 Takeaways

• Agents as Primary Interfaces: Agents will soon dominate customer interaction, unifying channels and raising expectations for service and sales.
• Process > Person: Productivity gains emerge not from "AI for jobs" but from optimizing and automating business processes.
• Messiness Over Elegance: Real-world agent systems thrive on messy, context-rich harnesses rather than neat engineering abstractions.
• Outcome-Based Business Models: Aligning pricing to agent-driven business results is both more aligned and more disruptive than usage-based pricing.
• Empowerment of Generalists: People who combine tech taste, infrastructure skill, and customer understanding—generalists with high agency—are increasingly leveraged by AI, upending traditional corporate specialization.
• Changing the SaaS Landscape: The future value center may shift from traditional “systems of record” (databases) to agent-embodied processes.
• Actions at the Board Level: Post-crisis leadership at OpenAI, with a board that thinks more about mission than margin.
• Predictions for 2026:
  • Mainstream adoption of agents (the “year of agents”),
  • AI-authored code as default,
  • At least one mainstream scientific breakthrough headline,
  • Further evolutionary surprises as organizations adapt.

For listeners: This episode analyzes and predicts how AI transforms both the visible (customer experiences) and invisible (organizational structure, pricing, software economics) workings of companies. It’s a must-listen for anyone building, managing, or investing in the future of software and AI.