Box co-founder and CEO Aaron Levie posted on X that there is “a huge opportunity for resourceful and entrepreneurial talent within organizations to go in and reimagine workflows for a world of agents,” according to a Forbes analysis by Josipa Majic Predin published Tuesday. Levie’s argument: coding agents like GitHub Copilot and Claude Code work out of the box because the domain is technical. For the rest of knowledge work, someone has to structure the unstructured data, map the processes, build the skills, connect the systems, and redesign the workflow for agent execution. That someone is the new high-leverage hire.
“For the rest of knowledge work there’s no way around this,” Levie wrote. “There’s really no way to shortcut any of this work. It has to be done by a person or people on the team.”
The Actual Work
Forbes details the labor Levie describes: structuring unstructured data so agents can access it, mapping existing processes, building skills and plans, connecting disparate systems, and redesigning process flows. Then the human governance layer: deciding where oversight is needed, how outputs get validated, how exception handling works. Box shipped Box Automate in September 2025 to address exactly this, described by TechCrunch as an operating system for AI agents that breaks workflows into segments, each augmented with AI or kept deterministic based on risk tolerance.
“The vast majority of enterprise workflows revolve around unstructured data, which actually represents about 90% of our corporate information,” Levie said in a post-earnings call cited by Diginomica. Legal review, M&A due diligence, contract management, clinical research. These have been manual because software has never been able to read a document and make a decision. Agents change that, but only after a human does the integration work.
Where the Money Is Going
The VC market is pricing in Levie’s thesis. Agentic AI startups raised $2.8 billion in H1 2025, led by autonomous workplace agents, according to AI Agents Directory data cited by Forbes. AI captured 53% of all global venture funding in Q1 2025, amounting to $59.6 billion, according to Crunchbase data cited by CVVC.
The capital is concentrating around platforms with enterprise adoption: Glean raised $150 million at a $7.2 billion valuation for enterprise knowledge agents. Cohere closed $500 million for enterprise agent deployments behind corporate firewalls. Sovereign wealth funds were involved in $46 billion of AI venture transactions in the first nine months of 2025, according to EY analysis cited by Forbes.
Jake Flomenberg of Wing Venture Capital told TechCrunch: “The companies growing fastest are the ones that identified a workflow or security gap created by GenAI adoption, then executed relentlessly on product-market fit.”
The Scale Argument
Levie expects 100 to 1,000 times more agents than people operating inside enterprise software systems, a number he shared at the Axios AI+ Summit in December 2025. That ratio breaks the per-seat pricing model that has underwritten SaaS for two decades and creates a new category of work: the person who designs, validates, and governs the agent layer.
“We are on day one of agent adoption in the enterprise,” Levie told GeekWire in May 2025. What comes after the pilot phase, if VC allocations hold, is rapid scaling into a talent market that does not yet exist at the scale enterprises will need.
Why This Matters for Builders
For solo operators and small teams in the agent ecosystem, Levie’s framing validates a specific bet: the most valuable skill in 2026 is not building the model or running the infrastructure. It is understanding a business workflow well enough to wire an agent into it. That is a skill that requires domain knowledge, not a computer science degree, and the VC market is funding the platforms that make it possible.