How is AI Different Than Other Technology Waves? (With Bret Taylor and Clay Bavor) [ACQ2]

Episode Overview

The episode features a discussion with Bret Taylor and Clay Bavor about how AI differs from prior technology waves, why adoption is accelerating, and what downstream effects could look like—especially for enterprise software.

They also explain their AI startup, Sierra, which builds customer-facing AI “agents”, and how Sierra’s pricing model is structured around successful task resolution rather than traditional SaaS metrics.

---

AI as a Qualitatively Different “Wave,” Not Just Better Software

The hosts debate whether AI is fundamentally transformational or merely “better software.” Their shared framing is that AI changes what is plentiful:

• Not just tools that boost productivity
• But intelligence becoming broadly available

They argue this shift can affect:

• Identity
• Labor
• Society

…often in uncomfortable transitional phases, similar to earlier historical technology transitions. At the same time, they expect greater access to expertise—for example:

• mental health guidance
• education
• legal/medical support

---

Why AI Adoption Is Unusually Fast

They compare adoption curves across major platform shifts:

• PCs enabled mass internet adoption
• Smartphones accelerated internet reach
• AI can now scale from “zero” to tens of millions faster because the distribution infrastructure already exists:
  • the internet
  • mobile connectivity
  • ubiquitous awareness channels like social media

They also discuss how shortened awareness/discovery cycles—driven by social platforms and widespread distribution—reduce friction for adoption.

---

The “Agents” Shift—and What It Implies for the Internet’s Business Model

Clay expects the term “agent” to stick because it matches user experience: autonomous digital actors.

They draw analogies:

• web → sites
• mobile → apps
• agents → autonomous actors that do the work

Sierra’s thesis is that AI agents will become more important than websites and apps because they will directly perform tasks, not just assist users.

Second- and Third-Order Market Effects (Internet Restructuring)

Bret and Clay focus on effects beyond the interface itself:

• AI may reshape the internet funnel: demand generation → discovery → fulfillment → transactions

• If personal agents manage interactions, human-driven marketing and decision journeys may shrink

• Demand could shift toward:
  • agent-to-agent workflows
  • agent-mediated decision and fulfillment paths

They emphasize uncertainty about the exact outcomes, noting that mobile and social already produced surprises.

---

Sierra’s Product: Customer Experience “Agents” That Take Actions

Sierra builds customer-facing AI agents for different parts of customer experience. They describe examples of agents completing complex tasks end-to-end, including:

• resolving support issues with no human involvement
• interacting with real-world systems, such as:
  • troubleshooting security panels
  • coordinating warranty replacements
  • triggering actions via satellite-based integrations

They highlight “agents interacting with the physical world” as evidence that capabilities extend beyond text generation.

---

Operating Philosophy: Build a “Machine” for Customer Outcomes

Internally, they treat the company like a system that produces “happy customers.” Their approach emphasizes fixing:

• the context that led to failure

…rather than only patching bad outputs.

They also mention using coding agents/tools (e.g., Cursor) to multiply engineering throughput, while still requiring:

• human oversight
• continuous process improvement

---

Sierra’s Pricing: Outcome/Resolution-Based

They argue traditional SaaS pricing doesn’t map well to what agents do. Agents complete jobs, rather than providing incremental productivity.

Sierra’s pricing is resolution-based:

• Customers pay when the agent successfully completes the task.
• Success is defined as human-not-required completion for specific cases.

Incentives Alignment

They claim the model aligns incentives:

• Sierra bears more operational risk
• Customers can evaluate clearer economic value by comparing:
  • what it would cost to solve with a human alternative
  • versus what Sierra costs

They also connect this to the historical evolution in ads—moving toward pay-per-conversion—and suggest it could reshape:

• enterprise software incentives
• procurement expectations

---

Hiring and AI Infrastructure: Where the Real Bottleneck Lies

They address talent competition indirectly, suggesting applied AI firms can succeed without competing for the same scarce frontier researchers.

They classify AI companies into three categories:

1. Frontier / foundation model labs (scarce top research talent + massive training capability)

2. Tooling / platform components (e.g., data labeling, vector databases, RAG tooling)

3. Vertical applied applications (e.g., Sierra and other agent builders)

They argue frontier model creation is constrained less by capital and more by:

• people
• massive infrastructure / capex (data centers, compute, highly specialized stacks)

They caution that foundation models can have short “useful lifetimes” due to rapid progress and competitive deterioration—while still acknowledging that frontier models may matter for quality and downstream advantages.

---

Leadership and Culture at Sierra

They contrast leadership in a smaller startup with earlier big-team leadership:

• In a smaller company, leaders stay close to details (pricing, contracts, engineering work)
• In the AI era, leadership also involves actively adopting AI tools so the company becomes a “poster child” for:
  • automation
  • workflow change

Bret emphasizes cultural empathy for how AI adoption shifts jobs. Clay emphasizes:

• excitement in rapidly iterating agent architectures
• deploying improvements quickly to real customers

---

Origin Story: Building Together

Their partnership is long-running:

• Both were Google associate product managers
• They were friends for years

The company started after late-2022 conversations (around the release of ChatGPT) when they shared conviction that:

• language models / agents would change everything

They present cofounding as an intentional partnership—focused on stress-sharing and collaboration rather than solo execution.

---

Presenters / Contributors

• David (Acquired host)
• Bret Taylor (guest)
• Clay Bavor (guest)
• Ben (speaker/questioner)
• Mark Benioff (mentioned)
• Sundar Pichai (mentioned)
• Marissa Meyer (mentioned)
• Sergey Brin / Larry Page (mentioned indirectly)