Submitted by Box
Content access, management and platform flexibility are emerging as dividing lines between AI leaders and laggards, according to a new report. State of AI in the Enterprise report From Box, which surveyed 1,640 IT decision makers in the US, UK, France and Japan. One of the key findings of the report is the speed of the shift: the overall share of organizations that describe themselves as pioneers or pioneers rose from 8% to 64% over the past year, while the share that called themselves early stage or not yet started fell from 53% to just 9%. Eighty percent of organizations reported a tangible return on their AI investment, defined in the survey as at least a 10% improvement, and more than half saw a measurable business impact within six months of project approval.
Box CEO Olivia Nottebohm says it’s more about how businesses now organize their use of AI than any technical breakthrough.
"We’ve moved from independent experience living at the individual level to systematized, integrated agent operations, agents that are in production and can be deployed in a repeatable manner." Nottebohm says. "This is where the impact comes from."
Why AI leaders achieve higher ROI than early-stage companies
The division between levels is a matter of implementation. Importantly, half of advanced companies reported an AI-based ROI above 25%, compared to only 11% of early-stage companies, a steady decline between the advanced (33%) and emerging (16%) tiers. But Nottebohm says the real differentiator was not whether companies used AI, but how tightly they integrated and managed it.
"What separates the advanced side is the operational muscle they’ve built: the right teams to deploy agents, formal governance to control them, and consistency in the content layer where those agents operate." he explains. "At an earlier stage, companies approach it in a more ad hoc, experimental way, letting people play with it without the same intent or structured design."
Content access is the biggest barrier to enterprise AI ROI
Content rather than model quality is the defining bottleneck of 2026. Ninety-six percent of organizations say agents need access to company-owned content, but only 36% have connected agents to trusted content in many use cases. It’s more a matter of trust than raw ability.
"We started this journey thinking that enterprise AI is about access to the latest model." Nottebohm says. "But now the question is whether agents have access to the right content and whether that content is protected, because these agents are only as good as the content they can reference and only as secure as the security around them."
Getting the content layer right has a second benefit beyond security, as it allows agents to work across previously isolated departments. While nearly a quarter of organizations point to fragmented data between systems, 24% struggle to integrate AI into existing systems, 21% say they lack adequate permissions and access controls, and 18% describe their content as too disorganized to make it generally accessible. Among the most mature organizations, 63% now see unstructured documents, contracts and reports as a competitive advantage rather than sitting in a digital filing cabinet.
Mitigation of general AI data exposure incidents
Nearly half of all organizations say they have already experienced an AI-related data breach. That number rises to 60% among advanced companies, which may face greater exposure from more agents and connected systems, but may be better equipped to detect it.
The share of organizations reporting established or advanced governance frameworks rose from 24% in 2025 to 73% this year, but real gaps in tools remain: only 39% have comprehensive visibility into sanctioned and unauthorized AI use, 34% have formal standards for how agents can access company data, and 27% are still describing their work. Nottebohm says these events act as a forcing mechanism, not a regression.
"Management used to be seen as something that slowed people down, but 93% of respondents told us that better management actually allows them to move faster." he explains. "This makes the scale of AI viable. Once the content is provisioned and highly authorized, you can run multiple agents in multiple processes and get a real multiplier effect."
One practical consequence of this change is that permission structures built for human workers are now being revised with agents in mind, a process most enterprises are only halfway through.
"Permits created two years ago should be revised." he explains. "Until fairly recently, people didn’t allow a document because of how an agent might use it, but now they think more about it. This leaves them with a whole corpus of unstructured data to go back and clean up or reauthorize."
This is part of a broader move away from management designed for people and towards management designed for agents from the start.
"Enterprises must transition from human workflow-enhanced management to management built specifically for agents," Nottebohm says. "This means tracking what an agent touches, whose permissions are applied, and what resources are used, all of which now shape how management is applied."
Enterprises should avoid locking into a single AI vendor
"Gone are the days of token-maxing," Nottebohm says. "Now it’s about the responsibility to deliver effective AI. Organizations want to use the least expensive model that meets the quality they need, not the most expensive model, because different model families leapfrog each other and companies want to maintain that choice."
This means that businesses are more reluctant to close than ever. Sixty-eight percent said they were worried about depending on a single AI provider, the average number of officially adopted AI tools rose to 3.3, and 79% now consider it important or critical for agents to operate headlessly, connecting directly to systems and APIs without a human interface between them.
This is a trend similar to the shift to multi-cloud infrastructure, and stems from a similar reluctance to give any vendor less than great negotiating power.
"An agile architecture is based on platform interoperability," Nottebohm says. "It runs on multiple models, is headless, and keeps every part of the AI stack changeable, so organizations don’t have to bet on which individual tool will win, and it’s part of a broader shift from the default to the biggest, most expensive model available."
Next steps for AI success
Over the next three years, enterprises should prioritize organizing, categorizing and cleaning unstructured content, proactively hiring and building teams around emerging roles, and adopting a hybrid token computing budget model where IT owns the core infrastructure and token budget. And at this point, it’s easy to accelerate quickly.
"You don’t have to start in early adulthood and work your way up," Nottebohm says. "If you build in governance, a content layer, and a multi-model system from the ground up, you can enter as a leading company and have the same impact."
Sponsored articles are content produced by a company that paid for the post or has a business relationship with VentureBeat and is always clearly marked. Contact for more information sales@venturebeat.com.
