90% of enterprise data is unstructured, and it's the biggest unlock for AI. Box COO Olivia Nottebohm breaks down what's separating AI leaders from everyone else. 👉 Download the full State of AI in the Enterprise report: https://www.box.com/state-of-ai?utm_source=youtube&utm_medium=organicsocial&utm_theme=icm&utm_campaign=SAI26 In just one year, early AI adoption dropped from 53% to 9%. Today, 64% of organizations describe themselves as advanced in AI, but only half of these leaders report significant ROI impact. So what's driving that gap? Box COO Olivia Nottebohm joins Nick Johnson to discuss findings from Box's second annual State of AI in the Enterprise report, based on a global survey of 1,600 IT decision makers. They explore why adoption has accelerated and what it takes to move from early-stage to leading edge. The two blockers nobody talks about enough Olivia identifies two root causes behind the ROI gap. The first is change management. LLMs are powerful, but paying for seats is not the same as having a programmatic approach to teaching people, modeling behaviors, and celebrating impact. The second blocker is unstructured data, which makes up 90% of enterprise content (documents, contracts, records). When AI agents work on uncurated or unpermissioned content, outputs aren't usable. Today, CIOs are refusing to open agents until their content is in order. What leading-edge companies do differently Mature organizations use AI agents for complex work that wasn't feasible before. Olivia shares two examples: a company that onboarded tens of thousands of employees in one week using metadata extraction after an acquisition, and a bank that cut client onboarding from a week to hours by automating document intake. Nearly two-thirds of leading-edge companies describe their unstructured data as an active competitive advantage. They have built a content layer that gives AI the enterprise-specific context it needs to deliver accurate answers. Governance as the accelerator While three-quarters of organizations say current governance slows agent rollouts, over 90% say better governance will let them move faster. Good governance is an accelerator: when people know agent data is secure, they trust outputs and move quickly. Model flexibility and the headless future Over two-thirds of leading-edge companies worry about locking into a single model. Mature organizations build architectures designed to win regardless of the LLM, including headless deployments where agents connect directly to systems. Box acts as the file system for AI, providing the context layer any agent can draw from. The workforce question Most organizations expect headcount to rise, but only 1 in 8 expect team structures to stay intact. New roles are emerging: agent operators, workflow specialists, and AI architects. Box itself has opened 13–14 new roles in the past 24 months. The narrative is shifting from compression to expansion. FAQs: Q: What does the Box State of AI in the Enterprise report measure? A: Based on a global survey of 1,600 IT decision makers, it tracks AI adoption maturity, ROI impact, governance challenges, and workforce changes. Q: Why are some companies seeing much better AI ROI than others? A: Olivia Nottebohm attributes this to change management (programmatic adoption, not just buying seats) and data readiness (curating, governing, and permissioning content before agents use it). Q: What is unstructured data and why does it matter for AI? A: Unstructured data (documents, videos, contracts, photos) makes up 90% of enterprise data. Accessing this content in a governed, trusted way is the biggest unlock for enterprise AI because it provides the specific context general models lack. Q: What does good AI governance actually look like? A: Governance should be an accelerator, not an inhibitor. Properly secured and permissioned content allows agents to move fast safely. The goal is governance built for an agent world, anticipating autonomous agent-to-agent interactions. Q: What are examples of AI doing work that wasn't feasible before? A: Examples include onboarding tens of thousands of employees in one week via metadata extraction after an acquisition, and cutting bank client onboarding from a week down to hours by automating document intake and routing. Q: How is AI changing the workforce? A: Headcount is expected to rise, but roles are shifting—only 1 in 8 teams will stay intact. New roles like agent operators, AI architects, and process flow designers are emerging. Box has opened 13–14 such roles in the past 24 months.

 59052      58