The co-founder of cloud storage company Box has rejected predictions that artificial intelligence agents will supplant enterprise software-as-a-service companies, instead forecasting a hybrid model combining traditional systems with AI augmentation.

Aaron Levie, speaking at a technology conference on Wednesday, argued that businesses will continue requiring deterministic systems for core workflows due to the risks associated with non-deterministic AI behaviour.

"Generally, once you have a business process, you want to be able to define that in, effectively, business logic with deterministic systems—just because the risk of that changing any given day is very high," Levie explained.

The executive cited examples of AI agents inadvertently leaking data or damaging production databases, suggesting organisations need separation between deterministic software components and non-deterministic AI elements—what he termed "church and state" divisions.

Levie envisions enterprise software futures where SaaS platforms handle core business workflows whilst AI agents operate atop these foundations, assisting with decision-making, workflow automation and process acceleration.

This architectural shift would fundamentally alter enterprise software economics, according to the Box chief executive. He anticipates organisations deploying 100 to 1,000 times more AI agents than human users, dramatically expanding the number of "users" accessing software systems.

Consequently, traditional per-seat pricing models would become unsustainable, requiring companies to adopt consumption-based and volume-oriented pricing structures aligned with AI agent usage patterns.

These changes present particular opportunities for startups building agent-first solutions rather than established companies attempting to retrofit agents into existing processes, Levie suggested. Smaller ventures lack entrenched business processes, enabling them to design workflows optimised for AI agent interactions from inception.

"We are in this window right now that we have not been in for about 15 years, which is—there's a complete platform shift happening in tech that's opening up a spot for a new set of companies to emerge," Levie stated, encouraging entrepreneurs to capitalise on the transition.

The perspective contrasts with more revolutionary predictions suggesting AI agents will entirely replace traditional enterprise software. Levie's hybrid vision acknowledges AI capabilities whilst recognising businesses' need for predictable, reliable core systems.

His emphasis on determinism reflects practical concerns emerging as companies experiment with AI agents in production environments. High-profile incidents involving AI systems behaving unexpectedly have heightened awareness of risks associated with delegating critical business processes to non-deterministic technologies.

The architectural separation Levie advocates—maintaining deterministic cores whilst enabling AI augmentation—mirrors patterns observed in other technology transitions where new capabilities complemented rather than replaced existing infrastructure.

Whether the industry evolves toward Levie's hybrid model or more transformative architectures remains uncertain. Competing visions include AI-native platforms designed without traditional SaaS constraints, potentially offering advantages over hybrid approaches requiring backward compatibility.

The pricing model implications Levie highlighted represent significant challenges for established SaaS vendors whose business models assume human users as the primary consumption unit. Transitioning to consumption-based pricing whilst maintaining revenue predictability requires substantial operational adjustments.

For startups, the platform shift Levie described could indeed create opportunities comparable to previous transitions like cloud computing or mobile that enabled new companies to challenge incumbents. Whether this materialises depends partly on how quickly enterprises adopt AI agents at scale versus continuing reliance on traditional software workflows.

Box itself must navigate these transitions, balancing protection of its existing SaaS business with necessary evolution toward agent-augmented architectures—a challenge facing numerous enterprise software companies as AI capabilities mature.