Box CEO Aaron Levie: AI Agents Won’t Kill SaaS—They’ll Supercharge It
In the whirlwind of AI hype, where doomsayers predict the end of enterprise software as we know it, Box co-founder and CEO Aaron Levie offers a grounded counterpoint: AI isn’t here to replace SaaS, but to layer on top of it, unlocking unprecedented value in workflows and business models. Speaking at TechCrunch Disrupt 2025, Levie envisioned a hybrid future where core SaaS platforms handle deterministic business logic, while AI agents automate decisions, accelerate processes, and expand total addressable markets (TAM) in ways traditional software never could.
Levie, whose company Box has evolved from a simple file-sharing tool in 2005 to a powerhouse in intelligent content management, draws parallels to past tech shifts like the cloud migration. Back then, enterprises resisted until the productivity gains became undeniable—much like today’s AI enthusiasm, which he compares to that initial cloud reluctance. “We’re at the very beginning,” Levie told the GeekWire Podcast in May 2025, emphasizing how AI agents—autonomous tools that execute tasks, follow instructions, and integrate with enterprise data—are reshaping interactions with content. Box’s own AI initiatives, like Box AI for content management and the recently launched Box Extract for data pulling from documents, exemplify this: agents now handle complex extractions from contracts or invoices, freeing humans for higher-level strategy.
At the heart of Levie’s thesis is the explosion of “agentic” workflows. Unlike passive chatbots, these agents actively complete tasks, but they thrive on rich context from SaaS platforms. “The models pack insane capabilities, but to get them into the middle of an enterprise workflow requires significant scaffolding, domain understanding, and context engineering,” Levie posted on X in July 2025. This isn’t sci-fi autonomy; it’s pragmatic integration. For instance, in coding—where tools like GitHub Copilot pioneered the shift—agents now generate code in chunks for human review, blending AI’s speed with oversight to avoid “mountains of slop” and security risks. Levie warns against overhyping fully autonomous agents, echoing Andrej Karpathy’s view that we’re years from “magical” superintelligence, but incremental value is already accruing through human-AI collaboration.
Enterprise adoption lags behind the buzz, Levie notes, with barriers like data fragmentation across systems and the need for workflow redesign. In a Wharton study he referenced in October 2025, 75% of companies reported positive ROI on AI, but smaller firms (<$2B revenue) are lapping larger ones in implementation, thanks to nimbler processes. “Demand from the business outstrips implementation,” he shared after hosting IT leaders in July, highlighting blurred department lines as AI enables cross-functional work. Security remains paramount: Box’s MCP server, now live in Anthropic’s Claude, ensures agents access only permitted data, preventing leaks in multi-tool ecosystems.
For U.S. enterprises, this AI-SaaS evolution means rethinking economics and operations. Traditional SaaS was seat-constrained—value tied to user count. Agents decouple that, scaling output exponentially: a 50-lawyer firm might now handle 500 contracts via automation, exploding TAM in underserved verticals like compliance or healthcare. Levie predicts “systems of intelligence” where agents farm out tasks across platforms, morphing jobs without eliminating them—think sales engineers crafting custom demos or developers tackling skunkworks projects. Politically neutral but economically seismic, this could widen productivity gaps between AI-adopters and laggards, pressuring mid-sized firms to upskill amid talent shifts. Lifestyle perks? Less drudgery in knowledge work, more time for innovation, though Levie cautions on training the “next generation workforce” to learn via AI-augmented ropes.
Levie isn’t blind to risks: brittle agents fail fast without context, and over-reliance could amplify errors. Yet, incumbents like Box, Figma, and Notion that rebuild AI-natively will thrive, while slowpokes perish. Startups have a shot in niches, but the trillion-dollar prize lies in agent interoperability and design—wild west territory for UIs that orchestrate humans and bots seamlessly. As Levie put it at Y Combinator’s AI Startup School in September: “AI is the most transformative shift yet,” blurring jobs while magnifying experts.
Looking ahead, Levie forecasts a decade of agents, with 2025-2035 as the “era of context” where engineering the right prompts, data flows, and interfaces determines winners. Box’s September 2025 launches—like AI Studio updates and Box Automate for drag-and-drop workflows—position it as a linchpin, integrating with OpenAI’s Agent Builder for multi-agent systems. The outlook? Explosive growth in SaaS categories once capped by labor scarcity, but success hinges on balancing determinism with AI’s creative spark. For enterprises, it’s not about surviving the AI wave—it’s about riding it to redefine work entirely.
By Sam Michael
Follow and subscribe to us for instant push notifications on breaking tech and AI news—stay ahead of the curve!
Aaron Levie AI, Box AI agents, enterprise SaaS AI, AI workflow automation, agentic AI future, SaaS business model change, Box CEO interview