Aaron Levie: Why Startups Win In The AI Era
Box CEO Aaron Levie discusses the paradigm shift from cloud to AI, arguing that AI will augment jobs by automating non-strategic work and create massive opportunities for startups to tackle previously unsolvable problems with unstructured data. He covers the evolution of SaaS business models, the 'build vs. buy' debate, and offers key advice for founders in the AI era.
A Unique Window of Opportunity
We are in a pivotal, but temporary, window of opportunity for building new companies. This period, which started about a year ago and will likely last another three to five years, is when the next hundred great companies will be founded. Unlike previous technology shifts, the enterprise is already convinced of AI’s potential, creating a fertile ground for innovation.
“Now is the moment. This window will end. In this window, between a year ago and 3 years or so, plus or minus from now, this is when the next hundreds of great companies will get started. So be ambitious because these windows don’t come… but for more than every 10 to 20 years.”
Lessons from the Cloud Wave vs. The AI Revolution
Box’s journey began in 2005, a time before the iPhone, Chrome, or widespread high-speed internet. The initial consumer-focused storage product pivoted to the enterprise, a move that proved fortuitous as it allowed the company to ride the massive growth wave of mobile and cloud adoption, which were creating a new IT architecture.
There’s a key difference between the cloud transition and the current AI shift:
- The Cloud Era: A significant challenge was convincing enterprises that the cloud was the future. Many were skeptical about security and reliability, requiring a major educational and sales effort to win deals. The push for the cloud was not coming from the CEO or heads of marketing; it was an IT-level decision.
- The AI Era: The conviction is already there. Decades of science fiction and the recent, tangible experience of tools like ChatGPT mean that business leaders are already bought-in. The head of marketing can personally experience AI’s power and immediately see its potential. The challenge is no longer about convincing them if they should adopt AI, but how to implement it in a way that is safe, reliable, and works with their proprietary data.
Unstructured Data: AI’s Biggest Enterprise Opportunity
Enterprise data exists in two primary forms:
- Structured Data: Information in databases, such as customer IDs, invoice numbers, and revenue figures.
- Unstructured Data: The vast majority of enterprise information, including contracts, documents, marketing assets, and presentations.
Historically, this unstructured data, often stored in platforms like Box, has been difficult to automate or analyze. You could query a database, but you couldn’t easily ask a question of your entire corpus of documents.
AI agents fundamentally change this. They can read, comprehend, and reason over unstructured content, transforming it from a static archive into a dynamic, queryable corporate asset. This unlocks immense value and creates a massive opportunity for startups to build AI agents for nearly every task and job function that interacts with this data.
AI and Jobs: Augmentation, Not Replacement
The common narrative that “AI is coming for our jobs” misses the reality of work inside large companies. A vast amount of employee time is spent on activities that are necessary but not strategic—finding information, manual data entry, and other repetitive tasks.
“You’ll read a lot of press that basically says AI is coming for our jobs… is most of the press is not inside of big companies seeing how much time we spend on useless activities that are necessary but not strategic.”
AI agents can automate this low-value work, freeing up human employees to focus on high-impact, strategic initiatives:
- Spending more time on breakthrough innovation.
- Engaging more deeply with customers.
- Proactively supporting clients instead of just being reactive.
Furthermore, every company has a backlog of strategic projects that are never started because the cost of human labor makes them economically unviable. AI agents make it possible to tackle this work. For example, an AI agent can translate an ad campaign into 100 languages for all market segments, a task previously too expensive to justify, enabling the company to grow faster and serve more customers.
While a massive incumbent might use AI for efficiency gains, for a 50-person startup, AI provides the leverage to operate like a 500-person company. This allows them to enter more markets, build features faster, and serve customers better, ultimately accelerating their growth and leading to more hiring.
New “Nouns and Verbs” for Startups
Just a few years ago, it seemed that the core “nouns and verbs” of both consumer and enterprise life—music, travel, payroll, CRM—had been solved by a wave of modern tech companies. This made it difficult for new startups to find entry points.
AI has created a new set of nouns and verbs. The opportunity is not simply to build “CRM with AI,” as incumbents like Salesforce will inevitably do that. The real opportunity lies in identifying categories of work, often done by professional services, for which no incumbent technology exists. AI agents can deliver these services via software for the first time in history.
“There’s a very very long list of things that software never did before that AI agents are perfectly primed to go do now. And that’s basically the opportunity set.”
Rethinking SaaS Business Models for AI Agents
The traditional SaaS business model, based on per-seat licensing, is fundamentally broken by AI agents. You’re no longer limited by the number of human lawyers at a company; you can sell the work of “unlimited lawyers.”
This necessitates a shift to consumption-based pricing, where customers pay based on the volume of work performed (e.g., per contract reviewed). The key to profitability is the value of the software built on top of the base AI models. Just as Box’s value is the software built on top of commoditized storage, an AI startup’s value will be the proprietary workflows, data connections, and unique agent capabilities it provides. This allows for high gross margins well above the raw token cost.
However, it’s crucial to pair a consumption model with a recurring subscription component to ensure predictable revenue and avoid customers who use the service for a one-off project and then churn.
The “Build vs. Buy” Debate and Core vs. Context
The idea that enterprises will simply use AI to build all their own software internally is unlikely. This is explained by the concept of “Core vs. Context.” Companies must decide what is core to their business (e.g., character IP for Disney) and what is context (e.g., their HR system). They invest their limited innovation resources in their core, while buying best-in-class solutions for context.
While a company could build its own HR system, it’s not a strategic differentiator. They would rather buy a solution from a vendor who is an expert, can be held accountable, and manages the long-term maintenance and liability.
Advice for Founders
- Read the Canon: To understand markets, disruption, and strategy, read Crossing the Chasm, The Innovator’s Dilemma, and Blue Ocean Strategy.
- Find a Tailwind: Don’t fight headwinds. Build in a market that is being fundamentally transformed by AI, where the technology changes the core economics or processes.
- Build a Great Team: Find at least one co-founder to go through the difficult times with.
- Be Ambitious: This is a rare moment in time. Use the next few years to go big and build something foundational.