Economist Podcasts – Boss Class: Interview with Bret Taylor (Sierra & OpenAI)

Date: January 29, 2026
Host: Andrew Palmer (The Economist)
Guest: Bret Taylor (Co-founder/CEO of Sierra; Chairman of OpenAI)

Episode Overview

This special episode of "Boss Class" features an in-depth interview with Bret Taylor—veteran tech executive, co-founder and CEO of Sierra (an AI customer service agent platform), and chairman of OpenAI. The conversation explores the explosive rise of AI-powered agents, the evolving landscape of enterprise software, challenges in deploying generative AI, regulatory and trust issues, and broader implications for knowledge work, customer experience, and jobs. Taylor provides insight from both the application layer (Sierra) and the foundational AI model level (OpenAI).

Key Discussion Points and Insights

1. The AI Agent Revolution

• Immediate Context:
  • AI-powered agents, particularly in customer service, are proliferating. The episode opens with a realistic, nuanced demonstration of a Sonos AI support agent, highlighting its strengths (patience, 24/7 availability), limitations, and natural tone.
  • Taylor’s vision: AI agents as the new customer interface, analogous to the invention and mass adoption of corporate websites in the 1990s.
    • Quote:

      "In 2026, every company needs an AI agent. And the vast majority of the digital interactions you’ll have with your customers will be via your agent." (Bret Taylor, 03:14)

2. The State of AI Integration—Early Days

• Parallel to Internet and Software Evolution:
  • We are in the "early innings" of AI agent adoption, similar to the early dot-com days where everything was bespoke, expensive, and clunky.
  • Few off-the-shelf AI agent solutions exist yet; most businesses are experimenting, often cobbling together tools and models.
  • In five years, Taylor foresees a robust ecosystem: for each business function, a dedicated AI agent can be easily purchased rather than built from scratch.
    • Quote:

      "Slowly but surely, we’re emerging with off-the-shelf solutions for the most important use cases in agents… You’ll have to go through this question of, do you want to incur that complexity and cost? You should probably prep yourself: you’re going to throw it out when a vendor is available." (Bret Taylor, 06:55)

3. Strategic Advice: Experiment or Wait?

• Should businesses jump in now or hold off for ready-made solutions?
  • Experimentation is crucial for organizational learning but carries opportunity costs if competitors outpace you.
  • AI is “inherently deflationary” (does more with less), making early adoption a competitive risk for laggards.
    • Quote:

      "There’s a risk to waiting too long because if you have a savvy competitor... what structural advantages might compound in the period where you’re waiting and your competitor is not?" (Bret Taylor, 08:24)

  • Experimentation also builds intuition that will be vital as tech—and business models—rapidly shift.

4. Imperfection, Hallucinations, and Guardrails

• AI is Not Infallible—Nor Are Humans:
  • The non-deterministic, probabilistic nature of large AI models complicates testing and robustness.
  • Rather than expecting perfection, design robust oversight and remediation processes—just as is done for human agents.
    • Quote:

      "Models are non-deterministic... That makes testing and the concept of robustness very challenging because you want to be able to say, ‘Will the model never do this?’ I’m not sure it’s possible to say that... People are imperfect as well." (Bret Taylor, 11:26)

  • Narrow domains (simple tasks like product returns) let designers engineer controls ("guardrails"); complex inference remains riskier.
    • "It goes from being a science problem to an engineering problem." (Bret Taylor, 12:56)

5. Regulated Industries & Risk Frontiers

• Surprising Adoption in Regulation-heavy Sectors:
  • Even banks and healthcare are experimenting, starting with low-stakes use cases (appointment scheduling, simple info gathering).
    • Quote:

      "I've been really pleasantly surprised how much regulated industries have adopted AI agents so far." (Bret Taylor, 13:54)

  • As organizations gain experience and regulation matures, higher-risk use cases will emerge.

6. Transparency & Customer Reactions

• Disclosing the Bot:
  • All Sierra AI agents openly identify as such and often admit occasional fallibility to build trust.
  • Taylor reports higher customer satisfaction scores with AI than human operators—patients appreciate the infinite patience and language flexibility.
    • Memorable Story:

      "An elderly man called because his receiver wasn’t working... spent more than 30 minutes on the phone talking to this AI and ended with, ‘Thank you, you’ve been a good robot’." (Bret Taylor, 15:56)

7. Humans in the Loop & Second-order Effects

• Balance Between Full Automation and Human Augmentation:
  • For complex or regulated tasks, human handoff (“humans in the loop”) remains essential.
  • Cost savings on simple queries can fund higher-quality human support for critical or nuanced cases; some clients have even ‘on-shored’ call centers again.
    • Quote:

      "The nice part about these AI agents... most people are opting into using them so you can end up having... higher quality, more expensive interactions with your people because you’ve essentially unlocked a bunch of budget." (Bret Taylor, 18:11)

8. Monitoring, Oversight, and Evaluation

• Multi-layered "Defense in Depth":
  • Real-time supervisor models monitor for hallucination and procedural errors, with problematic conversations automatically flagged for human review.
    • Quote:

      "You use a combination of AI and humans in the loop, using AI to help those humans that are in the loop so that they’re not just wasting their energy, but actually spending their time on problematic conversations." (Bret Taylor, 19:45)

• Key to Success:
  • Prioritize tangible business outcomes—like customer satisfaction—not just technical performance. Metrics should avoid being “gamed.”
    • Quote:

      "Specify a business outcome you’re trying to drive more than a technical outcome." (Bret Taylor, 21:56)

• Sierra's Model:
  • Outcome-based pricing: they charge only when the AI agent fully solves the customer’s issue.

9. Foundation Models vs. Application Layer: The Future of the Software Industry

• Will General Models Like ChatGPT Replace Enterprise Apps?
  • Taylor believes not; companies want “solutions to problems,” not just raw code or software generation.
  • Even as “the marginal cost of producing software is going down dramatically,” the real value is collective hardening, compliance, and continued updates—something best done by specialized vendors.
    • Quote:

      "You build it, you own it... I’m not sure just the cost of writing software is actually what you’re purchasing... My hypothesis is that you will buy agents that do purpose tasks." (Bret Taylor, 24:13)

10. Jobs, Reskilling, and Human Advantages

• Not Just Labor Savings—Better Customer Relationships:
  • The greatest value is in making customer engagement vastly more scalable and effective, not merely cutting costs.
    • Quote:

      "Our main value proposition is improving your sales and your relationships with your customers… Most CEOs are hired and fired based on growth." (Bret Taylor, 27:24)

• Impact on Work:
  • Taylor foresees unprecedented speed of job transformation, especially for knowledge work (“It’s like we’re all accountants and Microsoft Excel was invented last weekend”).
  • He is bullish on longer-term job creation, but acknowledges anxiety. Critical skills: adaptability, a beginner’s mindset, and willingness to learn AI tools.
• Advice to Workers:
  • Embrace reskilling. Even call center managers are “AI architects” now, driving value by applying domain knowledge to agent design.
    • Quote:

      "The people who are managing call centers are just as equipped... they can manage AI agents as well." (Bret Taylor, 31:39)

11. AI in Management: Personal Use

• Taylor’s Favorite Use Cases:
  • Using AI as a “creative foil” to critique ideas and as a time-saving tool for summarizing communication.
    • Quote:

      "I love using AI as a creative foil... I’ll use ChatGPT to critique it and find flaws." (Bret Taylor, 32:49)

• Desired Future Feature:
  • Seamless, conversational triage of email during commutes—Taylor believes it’s inevitable soon.

Notable Quotes and Timestamps

• AI agent goals:

  "If there’s anything I’m doing for society, it means you’ll never have to wait on hold again, which... is a very popular mission that’s way more important than AGI." (Bret Taylor, 02:44)

• On the current state of AI adoption:

  "We’re roughly in that era of AI and AI agents in particular, where everyone knows that agents are going to have a big impact… for a lot of use cases, there isn’t an off the shelf solution." (Bret Taylor, 04:45)

• Experimentation and intuition:

  "It is very important to experiment. As a consequence of things shifting so quickly." (Bret Taylor, 10:14)

• Guardrails in practice:

  "If you stop waiting for [AI] to be perfect and say it will be imperfect, do we have the technical and procedural controls in place to recognize when it is imperfect and remediate it?" (Bret Taylor, 11:50)

• Customer satisfaction:

  "The customer satisfaction scores of the AI agents built on Sierra are incredibly high and almost uniformly higher than the human interactions that preceded it." (Bret Taylor, 15:56)

• On software engineering disruption:

  "There’s no software engineer here in the Sierra offices that’s an expert in coding agents any more than any other... It’s like we’re all accountants and Microsoft Excel was invented last weekend." (Bret Taylor, 29:14)

Key Timestamps

• 00:30–02:44: Demonstration and analysis of a modern AI support agent (Sonos)
• 03:14: Taylor compares today’s AI agent moment to the early web days
• 04:45–08:00: The “early innings”—custom builds, lack of off-the-shelf AI solutions
• 08:24: Strategic implications of experimentation and waiting—why early adopters might win
• 11:26: Technical challenges: hallucinations, robustness, and the need for oversight
• 13:54–15:56: Regulated industries ramping up, customer reactions and trust
• 18:11–19:45: The "human in the loop" model and second-order effects in service orgs
• 21:56: How to properly measure and incentivize AI performance—business outcomes over technical metrics
• 24:13: Will foundation models displace software vendors? Taylor’s unique perspective as both app builder and model provider
• 27:24: On job loss, the future of work, and why the greatest value isn’t just cost-cutting
• 29:14–31:39: What individuals and teams can do to thrive—learning and adapting
• 32:49–33:59: Taylor’s personal use of AI for management and strategy; wishes for more seamless email triage

Memorable Moments

• Touching Story:
  An elderly customer thanks an AI bot for its patience after a long support call (15:56)—a symbol of how well-designed AI can sometimes outperform human empathy and endurance.

• Taylor’s “optimistic, mildly anxious” outlook:
  Taylor candidly discusses his own industry’s disruption and the emotional challenge for mid- and late-career professionals needing overnight reskilling (29:14–31:20).

• The “all accountants, Excel just invented” analogy:
  A memorable snapshot of how rapidly skills requirements are evolving for everyone, from junior coders to seasoned executives (29:14).

Takeaways

• AI agents are rapidly becoming the new customer interface, with major industry-wide implications.
• We are still in the transition phase—organizations must choose between experimenting now and waiting for mature, turnkey solutions.
• AI is not perfect and may never be; robust oversight, narrow domains, and human judgment remain key.
• Customer satisfaction is often higher with well-designed AI agents, especially when transparency is maintained.
• The main value for business is in enhancing customer relationships and experiences, not just cutting labor costs.
• The software and professional job landscape is set for the fastest change in history; learning AI is now a necessity at all levels.

Taylor’s final note is optimistic: the opportunity for career growth remains—provided that employees and organizations are willing to learn, adapt, and reimagine the future of work.