VIDEO
HIGHLIGHT
Log InSign up
A $4B founder on the one thing holding back every AI agent
SummaryTranscriptChat

A $4B founder on the one thing holding back every AI agent

The Evolution of Data and AI

Changes in Data Utilization

  • Agents now utilize data similarly to humans, contrasting with the past where computers could only handle structured data.
  • Historically, automation was limited to structured data in databases; unstructured data remained largely untapped until AI advancements.
  • AI represents a significant breakthrough by enabling automation for unstructured data, which constitutes over 90% of enterprise data.

The Role of Context in Data

  • The concept of "context" is emerging as crucial for understanding and utilizing data effectively within enterprises.
  • Organizations face challenges in providing AI agents with the necessary context to perform tasks efficiently, highlighting a major engineering problem.

Context Engineering Challenges

Importance of Context for AI Agents

  • Providing timely context is essential for AI agents to access relevant information and complete their tasks effectively.
  • Most discussions about AI adoption are fundamentally about how organizations manage their data and provide context rather than just technology capabilities.

Bridging the Gap Between Models and Real-world Adoption

  • There exists a gap between advanced model capabilities and real-world application due to bureaucratic processes that slow down adoption.
  • Key limitations include security concerns regarding access to data, user knowledge about systems, and the need for verifiable work outputs.

The Future of AI in Various Industries

Silicon Valley vs. Traditional Industries

  • Silicon Valley's digital economy allows faster adaptation of AI compared to traditional industries that rely on physical products and processes.
  • While sectors like life sciences may see rapid changes due to compute power, many industries will not adapt at the same pace due to inherent operational constraints.

Impact on Enterprises Using Box

  • Box serves various Fortune 500 companies by enhancing digital aspects of traditional businesses through improved collaboration and efficiency using AI tools.
  • Companies are leveraging Box’s platform for innovative applications that were previously impossible without human intervention, showcasing how AI can augment existing workflows rather than replace them.

Adapting Business Models for an AI-driven Future

Internal Changes at Box

  • Box is evolving its operating model by integrating more computational power into its business processes while maintaining core functionalities.
  • Identifying areas where increased computational capacity can enhance productivity is key; this includes onboarding employees faster or optimizing customer interactions through targeted messaging.

Headless Applications Strategy

  • As agent usage grows exponentially compared to human users, software must be designed API-first to accommodate this shift effectively.
  • A consumption-based pricing model may become necessary as agent activity could vastly exceed typical user engagement levels.

The Dynamics Between Labs and Applied Technology

Understanding the Relationship Between Hyperscalers and Applied Layers

  • Historical context shows that hyperscalers began with core infrastructure, gradually expanding into services. Despite expectations of them dominating technology, companies like Databricks and Snowflake have emerged as significant players.
  • The market cap for applied technology remains substantial, indicating a thriving ecosystem where numerous companies add value on top of existing infrastructure.
  • The applied layer is crucial as it bridges raw technological materials to customer implementation, addressing real-world needs such as compliance, security, and user experience.

Challenges in the Frontier Labs Approach

  • There’s potential for frontier labs to excel in specific verticals; however, a balanced ecosystem will likely emerge where both horizontal technologies from labs and specialized vertical applications coexist.
  • Companies focusing on vertical applications can provide tailored support to customers, ensuring effective integration of models into their operations while managing change effectively.

Economic Rationality in Applied Technology

  • The debate around model supremacy overlooks the necessity of additional components beyond just advanced models to achieve successful enterprise use cases.
  • Verticalized infrastructure stacks are essential for optimizing model efficiency based on specific use cases, especially given rising costs associated with advanced models.

Task Routing and Model Efficiency

  • A significant cost disparity exists between different tiers of models; understanding task requirements is vital for effective routing to appropriate models based on their capabilities.
  • Companies that grasp task intricacies will be best positioned to route tasks efficiently across various model tiers, enhancing operational effectiveness.

Evolving Principles for Founders in Today's Market

Adapting to New Realities in Software Development

  • Founders today face increased uncertainty; principles from past experiences should focus on building high-quality software that stands out amidst an influx of lower-cost solutions.
  • Engaging deeply with customers is becoming increasingly important as software development costs decrease. High-quality go-to-market strategies will become a new form of scarcity in this landscape.

Importance of Customer Engagement

  • Successful startups must prioritize direct engagement with customers rather than relying solely on product quality. Understanding unique value propositions is key to winning market share.
  • Investing in full-time employees (FTE), sales teams, and customer-centric approaches will be critical as software becomes easier to replicate but customer relationships remain complex.
Video description

Aaron Levie (Co-Founder & CEO of Box) has run Box through three platform shifts: cloud, mobile, and now AI. In this episode of Navigator, Paul sits down with him in Redwood City to talk about what's actually changing as agents take over knowledge work. Aaron makes the case that most "AI problems" are actually context problems. Agents need access to the 90%+ of enterprise data that's unstructured, and giving them the right information in a tiny window of time and space is the real unsolved challenge. He also breaks down why there's a growing gap between model capability and real-world adoption, why the applied layer will stay valuable even as frontier models get more powerful, and why founders building today need to treat go-to-market as the new scarce resource now that building software is cheap. CHAPTERS: 00:00 — Cold open: agents use data like people do 01:00 — Welcome to Navigator, intro to Aaron Levie and Box 01:53 — Redwood City as the new AI hub 02:20 — 20 years of Box, three tech shifts, and the throughline 04:07 — Data vs. context: why this plays into Box's hand 05:18 — Unstructured data is 90% of the enterprise and was the hardest to automate 06:36 — The real bottleneck: context engineering in a tiny window of time and space 07:46 — The model overhang: capability is outpacing real-world adoption 09:14 — Why diffusion doesn't speed up just because models get better 10:35 — Why coding got the fastest AI takeoff, and what the real world is missing 11:45 — Silicon Valley vs. the real economy: a divide that may never close 14:01 — How Box customers are actually using AI: speed vs. net-new capability 16:24 — Rebuilding Box's own operating model around agents 19:15 — Headless SaaS and what happens when agents outnumber human users 21:39 — Frontier labs vs. the applied layer: lessons from the hyperscaler era 25:41 — Token costs, model routing, and why the applied layer is positioned to win 27:45 — Advice for founders starting a company today: go-to-market is the new moat Browserbase gives AI agents a browser to operate in. Navigator is our series of conversations with the people building the agentic future.