← All case studies
B Box · ~$3B SaaS · Aaron Levie's company

What if every employee ran a team of agents?

Aaron Levie's bet: the next productivity unlock isn't a smarter chat — it's every knowledge worker quietly managing a fleet of AI agents in the background. Here's how Box is positioning for it.

100x
more agents than people (Levie's bet)
Background
where the real work happens
Manager
the new shape of every role
The problem

Chat is the wrong interface for most work.

Chat is great for quick questions. But most knowledge work isn't a single question — it's a long-running task: review 40 contracts, monitor 12 competitors, prep for tomorrow's board meeting. None of those fit in a chat window. They need agents that run in the background while you do other things.

What they're building

From chat in front of you, to agents behind you.

The shift Box is positioning for isn't a better assistant. It's a different shape of work — one where your job becomes managing the fleet, not doing the tasks.

Levie's bet on the shape of work
Layer 01 · Foreground
You + chat
One assistant in a chat window. The way most companies use AI today. Useful for questions, terrible for ongoing work.
Layer 02 · Background
Fleet of agents
Each agent owns a long-running responsibility: contract review, competitor watch, customer pulse, board prep. They run continuously. You don't babysit them.
Layer 03 · Manager
You, but managing
Your job shifts. You're no longer doing the work — you're directing a team of agents who are. You review outputs, adjust priorities, kill underperformers.
Where Box is investing

Box is investing in this shape.

Box Extract for document understanding, Box Automate for agent workflows, and a stated push to hire "agent engineers" who build and maintain these fleets. Aaron Levie has spent the last two years arguing publicly that this is what work looks like in 2027.

The big shift isn't AI assists humans. It's humans managing agents. Every knowledge worker becomes a manager of a small team. — Aaron Levie, CEO, Box (paraphrased from public talks)
What you can steal

Stop trying to make chat better. Start designing for fleets.

If your AI strategy is "roll out a chat tool and hope people use it," you're optimising the wrong layer. The companies that will pull ahead are the ones who treat AI like a team — multiple agents, long-running, in the background. Levie's been right about cloud, mobile, and remote work. He might be right about this too.

Sources

Aaron Levie on Latent.Space
Aaron Levie on AI & I (every.to)
Box market cap (macrotrends.net)