While Others Laid Off, Box Decided to Hire
When artificial intelligence began reshaping his business four years ago, Aaron Levie, the chief executive of Box, made a decision that ran against the prevailing current in Silicon Valley. While companies including Meta and Coinbase were announcing layoffs attributed to AI-driven efficiency gains, Levie chose to lean in — weaving AI into Box's products, embedding it into internal operations, and then asking a question most of his peers were not yet taking seriously: what new kinds of jobs does this create?
Box, based in Redwood City, California, makes software that helps companies store, manage, and work on documents and data. Its customers include more than 100,000 organizations, among them federal agencies and Morgan Stanley. The company went public in 2015 and has long taken a disciplined approach to hiring — its workforce grew just 20 percent from 2019 to 2022, a period when many peers doubled or tripled in size before facing brutal layoffs.
That restraint has now given Levie room to grow. Box expects to have more than 3,000 employees by early next year, up from 2,900 at the start of 2026 — a workforce expansion driven not by traditional growth logic, but by the proliferation of 13 new categories of jobs the company has created specifically because of AI.
"We ourselves are selling A.I. to our customers, so that's actually causing us to need to hire more people."
— Aaron Levie, CEO of BoxThirteen Titles That Didn't Exist Four Years Ago
The spark for Box's hiring expansion came from a practical problem. As the company began deploying AI across its own internal systems, it needed someone to wire everything together — to make its data, infrastructure, and AI tools work as a coherent whole. Last fall, Box posted a new job: senior director of AI, data and integration. It was the first of what would become thirteen new categories of roles, each one generated by the demands of operating a modern AI-infused enterprise.
The titles range from the strategic to the deeply technical. AI architects design the overall structure of how AI systems are built and deployed inside the company. AI solutions managers help clients understand which AI capabilities match their specific business needs. An AI platform leader oversees the infrastructure on which all of it runs. Each role reflects a gap that emerged when AI arrived — a gap between what the technology could do and what people in organizations actually knew how to do with it.
Selected New Roles at Box
Among the more unexpectedly human of the new roles is that of AI model evaluator. Sidharth Srinivasan, 23, joined Box full time last year after graduating from Stanford. His work involves testing the performance of AI models and helping customers determine which is best suited for which task — work that requires both technical judgment and an understanding of business context. "If I talked to myself two years ago, and I told myself I was working on A.I. evals, I would be like, what is that?" he said. "With any technological innovation, the type of work that you have to do to adapt to that is just slightly different."
How AI Makes Hiring More Justifiable, Not Less
Levie's argument for why AI is expanding Box's headcount rests on a logic that cuts against the conventional fear. The company is, simultaneously, a seller of AI products and a user of AI tools internally. Both dynamics push toward hiring more people, not fewer — though the reasoning is different in each case.
On the selling side, Box's AI products require human expertise to implement and support. Customers who want to use AI to automate contract review or document approval need engineers and solutions managers who can configure, customize, and maintain those systems. The more AI capabilities Box sells, the more human capacity it needs to deliver them. On the internal side, AI-driven productivity gains have allowed the company to justify roles that would previously have been too expensive. Box is now recruiting employees to market specifically to individual industries — a strategy it could not afford before AI made the underlying work more efficient.
"Now, you're hiring one or two to do the work of 10 because you can finally afford to do that work."
— Aaron Levie, CEO of BoxThe same dynamic applies to engineering. Box's pace of hiring software engineers has not slowed, Levie said, even as AI has become dramatically better at writing code. The reason is that AI agents — systems capable of performing tasks autonomously — have made each individual engineer more productive. A single engineer can now manage AI agents and accomplish what once required a team. That multiplicative effect makes each new engineering hire more valuable, which in turn makes hiring more of them more attractive. "Now that we can basically build way more features for our customers, it actually for us is attractive to have more engineers doing that," Levie said.
The Historical Parallel
Stephan Meier, a professor of business strategy at Columbia Business School, draws a direct comparison to the arrival of computers in workplaces during the 1970s and 1980s. That transformation produced entirely new departments, titles, and professional degrees — chief information officers, IT administrators, systems analysts — that no one had anticipated. Meier sees AI following a similar pattern, though he cautions that the new roles may not emerge quickly enough, or in the right places, to fully offset near-term displacement.
One Company's Experience Is Not a Universal Promise
The Times is careful to note what Box's experience does not prove. The growth of AI-related roles at companies like Box, or in cybersecurity (where demand for human expertise has surged to vet the explosion of AI-generated code), or at Google (which is recruiting more engineers for customer AI integration), is unlikely to compensate numerically for the scale of cuts underway elsewhere. Meta and Coinbase are not outliers — across the broader tech industry, the displacement of workers in the name of AI efficiency is real and continuing.
Meier, the Columbia professor, raises a further complication: it is not yet clear whether the new roles being created are permanent or temporary. As AI models become more capable, some of the integration and support work that currently requires human expertise may itself become automatable. The question Levie acknowledges openly — "when does A.I. slow down?" — is also the question on which the long-term durability of these jobs depends. If AI capability plateaus, the roles stabilize. If it continues accelerating, the jobs of today's AI evaluators and integration engineers may face the same pressure that today's AI is applying to yesterday's coders.
The Open Question
Whether AI-created jobs are permanent or a transitional category remains genuinely uncertain. The roles Box is filling today depend on the current gap between what AI can do and what organizations know how to do with it. How long that gap persists — and whether it closes faster than new gaps open — will determine whether the job creation story holds.
Box itself is not immune to these pressures. The company's stock has declined roughly 7 percent this year, weighed down by investor anxiety about whether AI will eventually allow enterprises to build their own software rather than buy it from vendors like Box. Levie's counter-argument — that third-party software will remain more secure and reliable than in-house AI builds — is a bet on the limits of AI self-sufficiency. It is a reasonable bet, but it is still a bet.
A Template Worth Watching for the Seattle Region
For the Responsible AI community in Seattle and Fremont, the Box story is worth examining not as a reassuring counterweight to fears about job displacement — the fears are real — but as a practical map of where new employment is actually forming. The specific roles Box is hiring for are not abstract futures. They exist today, they have salaries attached to them, and they require skills that can be learned: systems integration, model evaluation, AI-assisted process design, technical customer support for AI deployments.
The Seattle region, with its concentration of major technology employers and growing AI infrastructure, is likely to see versions of all thirteen of Box's new job categories emerge locally. Amazon, Microsoft, and the cluster of AI startups expanding in the area will face the same gap Box identified — between what AI can do and what their customers and internal teams know how to do with it. The workers who fill that gap will need both domain expertise in their existing field and enough familiarity with AI systems to evaluate, configure, and direct them.
Levie's formulation — "one or two doing the work of ten" — should not be read as comfort. It is an acknowledgment that the labor math has changed. But it also points toward the specific skills that remain valuable: judgment, integration, evaluation, and the ability to work productively alongside systems that are powerful but not autonomous. Those are learnable. And learning them, as Box's 23-year-old AI evaluator discovered, starts with being willing to take on a title that didn't exist two years ago.