From People to Agents

Office markets from San Francisco to New York have recovered nicely over the past few years, driven in large part by surging demand from AI companies. San Francisco is the epicenter. Tech demand accounts for 62% of all current San Francisco office demand.  29% of that is from AI.
 
Fundamentals are improving, valuations are rising, and investor interest is strong. All of this makes sense.  But every now and again, I find myself contemplating what might happen if agentic AI significantly changes how white-collar work is done.  As unpleasant as I find this thought exercise, I think it’s important to consider.
 
People to Agent Transformation
Agents are getting good at discrete white-collar tasks: drafting, analysis, coding, support, scheduling. If an agent can do a task as well as a person, at lower cost and without breaks, companies have an obvious incentive to use it. And once a company learns to deploy agents on one workflow, it will try the next one.
 
That alone does not empty buildings. Companies may grow into the productivity gains instead of cutting headcount.  Adoption inside large organizations is slow. Regulation, liability, and the need for human accountability may keep people in many roles.  And, historically, automation has changed the nature of work more than it has eliminated it.
 
But if agents keep improving and the pattern repeats up the hierarchy, organizations may become more agentic and less human. If superintelligence arrives and systems begin improving themselves, that shift could be sudden and sweeping, possible eliminating some industries, altogether.
 
My Struggle
My business is office buildings.  Office buildings exist so people can work. Agents don’t need an office. They need a data center.
 
This is where I struggle. I have heard the argument that AI will make starting companies so easy that opportunity will abound for otherwise displaced workers. But if agents are abundant, cheaper, and never sleep, why would new companies hire people? I don’t have a good answer, and I haven’t heard one that persuades me.
 
Two Risks Worth Monitoring
AI presents at least two distinct risks to the office market.  The first is cyclical: AI-driven office demand cools when capital tightens or valuations reset. The second is structural: the demand engine weakens because AI replaces the people who would have filled the space.
 
While no one knows how this will unfold over the coming decade, I can see a scenario in which the uncertainty gets priced into the market.  It could show up as more cautious buy-side underwriting.  Or we may see occupiers pushing for shorter and more flexible leases.  Most likely, both.  If we see an increased pacing of agents replacing humans, everyone will be forced to hedge against a future no one wants but which no one can fully ignore.
 
My Takeaway
It makes sense that some of us see an uncertain future as AI continues to proliferate.  But it’s too soon to draw any conclusions. What I know is that offices are built for people, and how many people, doing what kind of work, is an open question in a way it wasn’t five years ago.
 
For now, I’ll keep watching how AI affects business.  And I’ll continue to learn how it can sharpen my own craft. For my clients, I will increasingly push for decisions that hold up under more than one future.
 
I’d love to hear how you’re thinking about this.