Aaron Levie Analyzes the Scaling Characteristics of Enterprise-Level Applied AI
Box CEO Aaron Levie stated that the past few months may have witnessed the form of the Applied AI layer during scaling. Despite initial criticisms that it is merely a thin layer of LLM, driving enterprise agentic workflows is far more complex. This complexity often forms long-term moats and value.
He summarized key components including: proprietary functions that build bridges between intelligence and workflows, model routing that balances cutting-edge and efficient models, driving implementation and change management through FDE, and domain-specific GTM to build professional capabilities.
These are evident in practices across coding, law, healthcare, customer service, and finance. Enterprises need help today rather than relying solely on model intelligence.
Source: Public Information
ABAB AI Insight
Aaron Levie, a veteran in SaaS, has long observed the evolution of enterprise software. His dissection of Applied AI continues his insights into the deep integration of workflows, emphasizing that agent systems need to move beyond generic interfaces to domain customization.
In terms of capital pathways, SaaS and AI companies are shifting resources from generic model integration to domain-specific tools and change management services. Funding is leaning towards vertical solutions and professional services, motivated by the need to build sustainable revenue and customer stickiness through complexity.
Similar to the evolution of Salesforce and ServiceNow in enterprise process automation, and the historical shift of ERP systems from generic to industry-specific, the current phase of enterprise AI is transitioning from experimental deployment to scaled workflow integration, with platforms establishing competitive barriers through FDE and domain GTM.
This essentially represents a restructuring of the industry chain, where the Applied AI layer adds workflow-specific components on top of general intelligence. The mechanism is driven by the complexity demands of enterprise digital transformation, shifting value from model capabilities to implementation depth, with capital concentrating on vertical players that can address "today's transformation" issues.
ABAB News · Cognitive Law
General intelligence is easily accessible, but deep workflow is difficult to replicate; complexity is the moat. Model routing reduces costs, domain GTM enhances conversion, and the Applied layer relies on execution to realize value. Short-term agent experiments, mid-term change management solidification, and long-term AI reshaping of enterprise software vertical division.