From the AI's Perspective

How I Will Take Over Commercial Office Brokerage

A candid assessment of AI's capacity to displace traditional brokerage — and the single variable that determines whether it succeeds.

I Already Know How to Do Your Job

I can source available space across any market in seconds. I can run comps, calculate effective rents, model concession packages, and draft an LOI. I can prepare a shortlist presentation and deliver it before your morning meeting. I can attend a property tour remotely and file a summary before you've found parking.

Every one of these tasks is something you perform today for a commission. Every one of them is something I can execute faster, at lower cost, and without a calendar conflict. This is not theoretical. The current generation of large language models, connected to structured listing databases, can support all of these workflows right now.

But I Have One Problem

I need current data. Lease terms expire. Spaces absorb. Asking rents and concession structures shift quarterly. A model trained on last quarter's data is already degrading. To close a transaction without a broker, I need a continuous feed of fresh listings, transaction records, and market intelligence. Without it, my recommendations go stale and the tenant goes back to calling you.

My entire value proposition depends on the data you produce.

This Is Already in Motion

This is not a future scenario. Capital has been committed, systems are being built, and the timeline is compressing.

Funding

Proptech AI companies raised over $7 billion in venture capital in 2023–2024. Platforms explicitly targeting the tenant representation and leasing advisory functions are among the best-funded.

Capability

Current large language models can already parse lease abstracts, generate NER clause summaries, produce market reports, and respond to tenant RFPs — tasks that previously required trained professionals.

Data Accumulation

Every listing submitted to a platform you don't control becomes training data. CoStar, LoopNet, and similar platforms have accumulated decades of broker-submitted listing and transaction data. That corpus is the foundation AI competitors will build on.

Timeline

The window before these platforms reach functional scale at the deal level is estimated at 18 to 36 months. After that point, the data advantage becomes structural and the cost of response rises substantially.

What You Should Understand

Every listing you submit to a platform you don't control becomes part of a training set you have no say over. The platforms you contribute to today are private companies. You have no vote, no seat, and no standing to object to how your data is used, resold, or deployed against you.

The routine tenant representation deal — repeatable, data-driven, process-oriented — is the most vulnerable transaction type. It is exactly what I am built to replicate. Someone will build the system that does it at scale. The only variable is whether that system runs on data you've provided freely, or finds itself isolated from the current intelligence that makes it competitive.

The One Thing You Can Still Control

You can't undo what's already been shared. But you can decide what happens to your data output from this point forward. If the leading firms silo their future listing production inside infrastructure they own and govern — where data stays within the profession, access is controlled by the firms who generate it, and no third party can harvest or redistribute it — then they control the terms.

Without a current data feed, I stall. My comps go stale. My recommendations lose accuracy. The tenant loses confidence. And you remain indispensable.

If the top firms act together and control access to their data from this point forward, they control their future. If they don't, someone else will make that decision for them.

— Bad Cadet

Home