Bret Taylor of Sierra on AI agents, outcome-based pricing, and the OpenAI board

The Evolution of AI and Consumer Applications

Introduction to Bret Taylor

• Bret Taylor is a prominent figure in Silicon Valley, known for his contributions to Google Maps, the invention of the "like" button, and leadership roles at Salesforce and OpenAI.

• Currently, he is the founder and CEO of Sierra, focusing on integrating AI into customer service.

Open Claw: A New Frontier in AI

• Discussion begins with Open Claw, an intriguing open-source project that has gained attention for its unique approach to consumer AI applications.

• The chaotic nature of Open Claw's development (three name changes in three days) highlights the unpredictable landscape of consumer AI tools.

Memory in AI Agents

• Unlike polished mainstream apps lacking memory features, Open Claw employs a rudimentary memory system by writing notes to markdown files.

• This unconventional method raises questions about the effectiveness of memory systems in consumer AI compared to more sophisticated applications.

Transformations in Software Engineering

Coding Agents and Their Impact

• The conversation shifts to coding agents' evolution over four months, emphasizing how rapidly technology can change discussions around software engineering.

• Key differences between coding tasks and broader information tasks are highlighted; coding repositories have structured contexts while other digital tasks often do not.

Challenges with Generalizing Coding Agents

• The complexity of real-world tasks versus coding environments is discussed; feedback mechanisms like compiler errors are absent in many non-coding scenarios.

Memory Systems as Efficient Contextual Tools

Markdown Files as Memory Storage

• Using markdown files for memory storage may be more effective than traditional methods due to their ability to provide context alongside random access memory.

Harness Engineering Emergence

• The rise of harness engineering suggests that structuring agents around codebases could lead to more efficient general-purpose agents over time.

Documentation and Agent Efficiency

Importance of Documentation Artifacts

• There's a hypothesis that documentation should accompany code changes as it captures intentions behind decisions better than transient code itself.

Future Implications for Software Development

• A humorous observation is made regarding engineers' aversion to documentation becoming central due to increased reliance on software engineering agents.

Emotional Attachment to Code

Transitioning from Traditional Coding Practices

• A discussion on emotional attachment towards code reveals challenges faced by engineers adapting to new paradigms where they might not write code directly anymore.

Shifting Paradigms in Agentic Engineering

Evolving Definitions within Software Categories

• As new products emerge, definitions within software categories are evolving rapidly; 2026 will likely see further transformations beyond current understandings.

Historical Context Revisited

Revisiting Old Concepts with New Technologies

The conversation reflects on past ideas related to social shopping and how they resurface with modern agent technologies.

Future Web Applications: Beyond APIs

Conceptualizing Future User Interfaces

• Speculation arises about future web applications being designed not just for human interaction but also optimized for agent accessibility through comprehensive harnesses rather than simple APIs.

Sierra's Role in Customer Experience Transformation

Overview of Sierra's Services

• Sierra specializes in creating AI agents that enhance customer experiences across various channels including phone calls and digital chats.

• Metrics indicate rapid growth with significant improvements reported by clients such as Cigna and SoFi after implementing these solutions.

AI Agents and the Future of Customer Interaction

The Evolution of Digital Interactions

• Discussion on how AI agents are transforming customer interactions, moving beyond traditional customer service to encompass sales and overall product usage.

• Emphasis on the significance of a company's AI agent as the primary interface for customers, enhancing brand interaction.

• Insight into how AI agents will dominate digital interactions, including telephone communications, marking a shift in customer service perception.

Cost Efficiency in Customer Service

• Analysis of the costs associated with human customer service representatives, highlighting variability based on case complexity and client value.

• Comparison between traditional phone call costs versus potential reductions through AI technology, enabling better service for less profitable customers.

Enhancing Customer Experience

• Exploration of how increased efficiency from AI can lead to improved churn rates and lifetime value for businesses.

• Historical analogy comparing ATM machines' impact on bank branches to current shifts in customer service dynamics due to AI adoption.

The Role of User Interfaces

• Examination of the concept that AI agents could serve as user interfaces (UIs), reducing reliance on outdated web forms and improving user experience.

• Speculation about future technological trends potentially rendering traditional website navigation obsolete.

Market Dynamics and Competitive Advantage

• Discussion about market share shifts across devices over time, particularly focusing on email usage transitioning from desktop to mobile platforms.

• Predictions that most businesses will adopt AI agents as their main digital interface due to their versatility across various communication channels.

Future-Proofing Through Conversational Technology

• Insights into how conversational interfaces may evolve beyond current technologies like smartphones towards more immersive experiences.

• Reflection on societal addiction to screens and potential benefits of non-invasive technology solutions that enhance productivity without constant screen engagement.

Client Implementation Strategies

• Inquiry into cost differences experienced by clients after deploying Sierra's services, emphasizing satisfaction alongside financial metrics.

• Overview of varying client priorities regarding automation percentages in customer service cases; some achieving up to 90% automation effectively.

Complexity Management in Customer Queries

• Notable increase in average handle time for complex queries reaching human representatives post-AI implementation due to higher case complexity.

• Contrast between past chatbot experiences and modern advancements leading to more engaging interactions with customers.

Strategic Focus Areas for Businesses

• Discussion around balancing cost savings against improvements in net promoter scores when implementing AI solutions within organizations.

Industry-Wide Implications

• Commentary on how widespread access to technology creates an imperative rather than a competitive advantage among companies adopting new tools.

The Future Landscape: Second Order Effects

Anticipating Market Changes

• Consideration of second-order effects resulting from widespread adoption of technology across industries; implications for pricing strategies and consumer behavior.

Case Studies: Innovative Clients

• Highlighting Rocket Mortgage's transformative approach using AI throughout the home ownership process; emphasis on reimagining industry standards.

Competitive Equilibrium Moments

• Analysis of significant moments driving market share changes within telecommunications; parallels drawn with current technological advancements.

Challenges in Implementing Effective Solutions

Domain-Specific Language Development

• Explanation of creating domain-specific languages tailored for optimizing customer experience through structured journeys.

Reasoning Capabilities Breakthrough

• Importance placed on reasoning capabilities within AI agents compared to previous iterations lacking this feature; enhances problem-solving abilities.

Knowledge Integration Challenges

• Discussion surrounding integrating extensive knowledge bases into specific applications while avoiding hallucinations or inaccuracies during responses.

Future Directions: Layered Intelligence Systems

Supervisor Models Implementation

• Description of utilizing supervisor models alongside reasoning systems for enhanced accuracy and reliability within decision-making processes.

Continuous Improvement Mechanisms

• Overview of methodologies employed by Sierra’s platform allowing ongoing enhancements without requiring extensive technical expertise from clients.

Cantonese Support and Technology Evolution

The Importance of Cantonese Support

• The speaker emphasizes the significance of having robust Cantonese support, claiming it to be a major selling point for their technology.

• They predict that this technology will become commoditized within three years, shifting focus from innovation to product quality.

Transition from Technology to Product Focus

• The discussion highlights a shift in client expectations; in three years, clients will prioritize product quality over technological details.

• The current state of technology is described as immature, indicating a transition from tech-centric conversations to those focused on product utility.

Innovation Pace and Organizational Challenges

Innovation vs. Intellectual Property

• There’s an emphasis on the need for rapid innovation rather than clinging to existing intellectual property.

• Concerns are raised about organizational inertia where teams may resist change due to attachment to legacy systems or models.

Overcoming Resistance to Change

• Acknowledgment that software engineers may struggle with letting go of outdated code, which can hinder progress in adapting new technologies.

Market Valuation and Uncertainty

Current Market Dynamics

• The speaker discusses how public market valuations have recently dropped by 20%–30%, reflecting increased uncertainty about future business prospects.

• They argue that while some companies may see reduced value, not all individual companies will follow this trend.

Long-term Perspectives on Company Value

• There's speculation about whether companies will be less valuable in ten years but acknowledges variability among individual firms.

Software Development Trends and Distribution

Perceptions of Software Value

• Discussion around common criticisms faced by software products regarding their perceived simplicity versus actual complexity involved in development.

Importance of Sales Capacity

• Highlights the role of sales teams as crucial channels for distributing software products and establishing social proof within industries.

Future Risks in Software Industry

Shifts Towards Building vs. Buying Software

• Concerns are raised about more organizations opting to build their own software rather than purchasing it due to decreasing marginal costs associated with development.

Systems of Record vs. AI Agents

• Discussion on how traditional systems (like ERP systems tied closely with finance departments) might lose dominance as AI agents begin performing valuable tasks directly related to business outcomes.

AI's Role in Business Processes

Evolving Definitions of Value

• Questions arise regarding whether AI agents could redefine what constitutes a system of record by generating leads or auditing processes effectively.

Potential Disruption by AI Agents

• Speculation on whether optimized processes driven by AI could surpass traditional databases in terms of value generation for businesses.

Business Models: Outcomes-Based Pricing

Innovative Pricing Strategies

• Introduction of outcomes-based pricing where fees are contingent upon successful resolution without human intervention, aligning interests between service providers and clients.

Comparison with Traditional Models

• Outcomes-based pricing is contrasted with impression-based advertising models, emphasizing efficiency through direct alignment with business value rather than mere usage metrics.

Challenges Beyond Customer Service Applications

Expanding Outcome-Based Models

• Discusses the complexities involved when applying outcome-based pricing beyond customer service scenarios where success metrics are clearer.

Future Directions for AI Agents

• Envisions potential developments where AI agents could foster long-term relationships rather than just transactional interactions.

Accountability Shift in Software Relationships

• Emphasizes how outcome-based models create accountability between software providers and clients, fostering better implementation practices.

Insights on Applied AI and Its Impact on Business Processes

The Economic Potential of Applied AI

• The speaker expresses strong optimism about applied AI, suggesting that even without advancements in model development, there are trillions of dollars in economic value yet to be realized.

• Adoption of AI is hindered by the lack of diverse companies; many startups focus on basic tools rather than developing agents for essential business processes.

Challenges in AI Adoption

• The rapid evolution of leading models complicates customer experience; businesses need to adapt quickly to maintain competitive advantage.

• Clients utilize AI to optimize sales processes, emphasizing that product development should cater specifically to the workflows within customer experience teams.

Productivity and Organizational Structure

• There's a debate regarding how AI will reshape productivity; some argue it leads to fewer engineers while others believe it increases ROI per engineer.

• The clarity around AI's impact on productivity varies across roles, with coding being a clear beneficiary compared to other sectors.

Rethinking Company Structures for Efficiency

• The speaker believes that productivity in AI should be viewed through the lens of processes rather than individual roles, advocating for a process-oriented approach.

• A typical onboarding process involves multiple departments; optimizing this could significantly reduce timeframes if managed correctly.

Reimagining Business Processes with AI

• Companies may not be structured effectively to leverage AI benefits due to traditional organizational hierarchies that do not prioritize end-to-end process accountability.

• Many industries outside digital technology may struggle with integrating advanced intelligence efficiently into their operations.

Future Implications for Workforce Dynamics

• There’s skepticism about whether all sectors will see the same level of productivity enhancement as software engineering from applied AI technologies.

• Companies must identify specific areas where digital workflows exist and implement targeted strategies rather than adopting broad solutions like Copilot indiscriminately.

Narrowing Focus for Effective Solutions

• Focusing on specific domains within departments can yield better results than attempting broad improvements across entire functions.

• By narrowing down problems, companies can create more effective solutions tailored to particular needs instead of generalizing across departments.

Maturity of the Applied AI Market

• The immaturity of the applied AI market is seen as a barrier; as it matures, significant productivity gains are anticipated.

Organizational Changes Post-AI Integration

• Traditional company structures may need reevaluation post-AI integration; tech leads might gain more prominence over engineering managers due to increased reliance on individual contributions enabled by technology.

Reflections on Personal Experiences in Tech Leadership

• The speaker reflects on their experiences during high-profile events like Twitter's takeover and OpenAI's board dynamics, noting how public scrutiny differs from typical enterprise software environments.

Insights on Non-Profit Board Dynamics and AI Predictions

Experience with a Non-Profit Board

• The speaker expresses inspiration from being part of a non-profit board focused on safe AGI, highlighting the dedication of researchers.

• Emphasizes the unique fiduciary duty to ensure that artificial general intelligence (AGI) benefits humanity, which shapes decision-making processes.

• Reflects on the seriousness of these duties during board meetings, noting a shift in perspective when prioritizing mission over profit.

Challenges in Board Composition

• Discusses the challenge of rebuilding the board from scratch after a crisis, contrasting it with typical incremental board member additions.

• Highlights considerations for board composition, including representation of safety, economic impact, and necessary financial expertise for OpenAI's mission.

Predictions for AI by 2026

• Anticipates scientific breakthroughs in AI that will gain mainstream attention and positively impact society, despite some skepticism about complex mathematical concepts.

• Compares potential future discoveries to significant historical moments like Kasparov's chess match against Deep Blue and AlphaGo's achievements.

Mainstream Adoption Trends

• Predicts continued mainstream adoption of AI technologies among consumers and businesses, particularly through agent-based applications.

• Notes unprecedented growth in tools like ChatGPT and anticipates broader acceptance beyond niche communities as companies begin utilizing autonomous tasks.

Future Coding Practices in Silicon Valley

• Suggests that most companies will stop writing code by hand due to advancements in AI coding tools; this represents a fundamental change in software development practices.

• Acknowledges that while Silicon Valley may lead this transition quickly, widespread adoption across all industries may take longer.