Box CEO Aaron Levie is making the case that AI agents will multiply the number of workers using business software rather than thin out the humans behind it, arguing for a 10x expansion in system users within three years. Speaking on the debut episode of the Platformer podcast, Levie predicted that the ratio of human to agent activity on enterprise software will flip from roughly 90/10 today to 90/10 in the opposite direction by 2029. The pitch lands as tech companies announced nearly 46,000 layoffs in March 2026, the worst single-month total in more than a year, with a rising share of executives citing AI as a factor.
Levie's argument cuts against the prevailing SaaS-apocalypse narrative pushed by investors and founders who expect Anthropic-style agents to gut seat-based software pricing. Box stock has held up materially better than most SaaS peers over the past year, and Levie is using the moment to articulate why he thinks the seat model survives. His framing: the human side of the equation does not drop by 90% — instead, agents arrive as a new workforce layered on top, making the underlying systems more valuable, not less.
"The big idea is that in the future, if today maybe 90% of activity on this software is humans interacting with the interfaces, probably three years from now it'll be 90/10 the other direction," Levie said. He frames the open debate for the industry as whether that flip represents a 90% reduction in human seats or a 10x increase in total workers driven by agents running in parallel.
Key facts
- 01Box CEO Aaron Levie predicts a 90/10 flip in software interaction within three years, with agents handling 90% and humans 10%.
- 02Tech companies announced nearly 46,000 layoffs in March 2026, the worst single-month total in more than a year.
- 03Levie argues the shift represents a 10x increase in workers using systems via agents, not a 90% cut in human seats.
- 04Levie says vibe-coded output covers the first 80% of a job, while domain expertise lives in the remaining 20%.
- 05Amazon plans to hire roughly the same number of software engineering interns in 2026 as in recent years.
The second pillar of Levie's case is what he calls the durability of the last mile. Vibe-coded prototypes and LLM-drafted documents, he argues, cover the first 80% of any professional job. The remaining 20% is where domain expertise, judgment, and value creation actually live — and that's the part current models struggle to automate.
“If today maybe 90% of activity on this software is humans interacting with the interfaces, probably three years from now it'll be 90/10 the other direction.”— Jaeden Schafer
"If you or I go and vibe-code something, we think we've replaced the engineer, replaced the accountant, replaced the lawyer," Levie said. "But then you actually look — that was the first 80% of the job. The extra 20%, it turns out, is all the value creation of that profession. All the expertise and domain knowledge is in that last 20%, not the text that got generated."
The data points pull in both directions. Anthropic's Economic Index shows AI-related work conversations climbing across nearly every white-collar profession, and entry-level roles — the rung most exposed to LLMs — are showing the earliest disruption signals. Google DeepMind workers in the United Kingdom recently voted to unionize. At the same time, Amazon plans to hire about the same number of software engineering interns in 2026 as in prior years, and openings for software engineering roles sit at three-year highs.
On business model, Levie acknowledges seat-based pricing faces real pressure in categories where headcount falls, but argues most enterprise software does not sit in that bucket. Looking at internal Box consumption, he says there are few categories where he can build a case to cut seats, but many where agent-driven use cases multiply on top of existing licenses. The pricing question becomes how vendors charge for that agent layer, not whether the human layer disappears.
The deterministic core of enterprise software — ERP at a Fortune 500, CRM at Salesforce, document governance at a pharma company or bank — stays intact in Levie's view. What changes is the interaction surface: agents pulling data, executing workflows, and talking to systems on behalf of users. That shift, he argues, raises the value of the data plane and the guardrails around it rather than collapsing demand for the platforms holding the data.
Levie's argument has obvious self-interest baked in — Box sells the document plane he's describing — and his 90/10 prediction is a three-year forecast with no benchmark to hold him to. The counter-case is straightforward: if agents can handle 90% of software interactions, the buyer's logical next question is why they need the same number of human seats underneath. Anthropic's own data on white-collar AI usage suggests the disruption is already concentrated in entry-level work, which is where the seat-multiplier theory faces its earliest test.
The Levie framing matters because it's the cleanest articulation yet of the bull case for incumbent SaaS in an agentic world: agents are net-additive demand, not substitution. Whether that holds depends almost entirely on how SaaS vendors price the agent layer and how quickly the last-mile 20% gets eroded by the next model generation. For Box and its peers, the next 36 months will determine whether seat-based revenue compounds with agent revenue on top, or whether the seat curve bends down faster than the agent curve bends up.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
