The conventional wisdom around AI agents suggests a leveling of the playing field. After all, sophisticated prompts can now deliver expert-level results to anyone. Shouldn’t this mean that every organization is now far closer to equal?
Surprisingly, early enterprise data reveals the opposite. Organizations with deep human domain expertise extract substantially higher value from AI agents than those attempting to bypass fundamental competence. This pattern isn't a limitation of the technology. It reflects how knowledge work actually functions.
AI agents don't replace expertise. They amplify what you already know how to do.
Key insights:
- Organizations achieve the largest AI productivity gains when domain experts direct and evaluate agents, because experience (rather than simple prompt writing) is the driver for exceptional outcomes
- Realizing the full potential of AI requires redesigning processes and embedding AI into core workflows, which in turn depends on deep domain knowledge to set tolerances, oversight, and integration
- The most competitive companies upskill existing experts to combine domain mastery with AI fluency, since AI amplifies value only when layered onto real expertise
Domain knowledge drives the best AI outcomes
When organizations deploy autonomous agents to execute complex background tasks, the results demand skilled evaluation and integration (in other words, a smart human in the loop). Human domain experts can immediately see what succeeded, what failed, and how to refine prompts for better outcomes. They can distinguish acceptable output from exceptional results — a judgment that requires experience no prompt can simulate.
The biggest returns from agents are going to come to those that already have some existing expertise in that particular field
Box's State of AI in the Enterprise report confirms this. Companies that achieve productivity gains of more than 50% approach AI fundamentally differently than the low performers. AI leaders deploy AI strategically where existing expertise can leverage its AI capabilities rather than expecting AI to compensate for missing skills.
Good communication is a critical skill in the mix here. As Jon Allen, CIO of Baylor University, says, "Some of the greatest prompts I've seen have come from people who are not technologists. They come from folks who really understand language and how to ask and instruct things."
Effective AI use requires both mastery of communication and the skill to break a complex, large-scale task into smaller, more manageable, well-defined sub-tasks. These are competencies experts develop through years of practice, not tools novices can access simply through better AI interfaces.
Process redesign demands specialized knowledge
Many organizations simply bolt AI onto existing workflows, then wonder why productivity gains are disappointing. True AI leverage requires rethinking entire processes — work that depends on deep domain understanding.
"If you want to realize the promise of AI, especially with workflow automation, you cannot just add AI on top of your existing applications," explains Box Senior Director of Product Management Nirmal Ganesh. "You have to bake AI into the core experience to see productivity gains."
Building AI-native workflows requires understanding not just which tasks agents can handle, but how human oversight should be structured around their capabilities. Experts understand critical decision points, acceptable error tolerances, and downstream implications of failures. For process optimization to be successful, you have to know what to optimize for — knowledge that comes from experience, not experimentation.
The premium of upskilling workers
While smart, experienced domain experts are key to AI success, technology savvy is certainly a key skill in the AI-first business world.
Box research makes the investment pattern explicit: Companies reporting high AI productivity gains are heavily upskilling existing workforces. The majority of companies Box surveyed address AI skills gaps through training current employees rather than hiring new talent. Organizations recognize that AI fluency layered onto domain expertise creates more value than AI fluency alone.
Stanley Toh, Head of Enterprise End User Services at Broadcom, observes this skills transformation across his 40,000-employee organization. "The skillset is very different now," he notes. "The CIO and the IT leadership must look at different kinds of skillset to support the future IT moving forward in an AI world, including upskilling the talent that has AI expertise. You may also need data scientists — because you have so much data, and need to know which AI tool to use to make sure that your data is relevant."
In other words, professionals still need deep domain understanding. But they now also need the facility to effectively direct AI agents, evaluate output critically, and integrate results into larger systems — advanced capabilities that assume baseline technological competence.
Competitive Implications
Agentic AI in the field of software development has advanced fast, largely because developers understand their workflows intimately and are able to iterate rapidly on improvements. Over time, that pattern will replicate across every vertical, with domain experts leading adoption in each field. Levie identifies the pattern clearly: "AI coding agents offer the best glimpse into what the future of agents will look like in many other fields of knowledge work."
The competitive dynamics are straightforward. Organizations investing in their expert talent — providing AI tools and teaching effective leverage — will outpace competitors who merely expect AI to compensate for capability gaps. Box research shows early adopters achieving 37% average productivity lifts, with 60% of companies expecting AI transformation within two years. The window for establishing advantage is compressed.
"Don't lose your skills," Levie warns. The productivity gains from AI are substantial and real. But they flow to those who bring expertise to the table. The path forward isn't choosing between mastering your craft and adopting AI. It's recognizing that one makes the other valuable — and the organization that develops both will define competitive advantage in the AI era.
