Build vs Buy vs Wait: A Framework for AI-Era Brokerage Tooling
Build vs buy vs wait is a decision framework for brokerage technology investment. It evaluates any proposed tool against three dimensions: the half-life of the underlying capability, the defensibility of the workflow it touches, and the cost of being wrong. In the AI era, the third option (wait) is the correct answer far more often than operators realize.
Every brokerage I've talked to in the last eighteen months has a technology decision stuck in committee. Someone wants to build. Someone wants to buy. Nobody wants to wait because waiting feels like losing. So the decision gets framed as build vs buy, a vendor gets picked, an integration budget gets signed, and six months later nobody uses the thing.
That failure mode has a cause. The frame is wrong.
In the AI era, waitis the third option, and it is the correct answer far more often than operators admit. The capabilities you're about to buy or build will, in most cases, be delivered natively by the platforms your agents already use, or by generalist AI agents, within 12 to 24 months. The question is which category a given decision falls into, and there is a framework for that.
The three variables
Every technology decision at a brokerage should be evaluated on three dimensions. Not two. Three.
- Half-life of the underlying capability.How long until this capability is commoditized by a platform (Microsoft, Google, your CRM, Zillow, MLS, or a general-purpose AI agent)? Under 12 months means the build is a waste before it ships. 12 to 36 months means it's a bridge. Over 36 months means it's infrastructure.
- Defensibility of the workflow it touches. Is this workflow a source of operating leverage unique to your brokerage, or a table-stakes function every brokerage runs the same way? Differentiated workflows can justify builds; commodity workflows cannot.
- Cost of being wrong.What does it cost in sunk capital, opportunity cost, and agent trust if the decision needs to be reversed in 12 months? A failed vendor costs the license fee plus the training effort. A failed build costs double that plus the organizational scar tissue of "we built a thing nobody uses."
The 2×2×2 (but really just a decision table)
Most brokerage tech decisions collapse cleanly into five outcomes once you apply the three variables. The table below is the working version I use:
| Half-life | Defensibility | Cost of wrong | Call |
|---|---|---|---|
| < 12 mo | Any | Any | Wait. Don't build and don't buy. |
| 12–36 mo | Commodity | Low | Buy. Rent the bridge. |
| 12–36 mo | Commodity | High | Wait or buy on short contract. Avoid lock-in. |
| > 36 mo | Differentiated | Low | Build small, buy around it. |
| > 36 mo | Differentiated | High | Build. This is the rare true-build case. |
Notice how small the "build" surface actually is. For the median residential brokerage, there are maybe three to five workflows that clear all three bars simultaneously. Everything else is buy, wait, or a combination.
Why "wait" is undervalued
Waiting feels like inaction, and inaction gets punished in operator culture. That instinct is calibrated for a world where capabilities compound only if you build or buy them. It is miscalibrated for a world where the largest platforms in your stack are shipping new capabilities every quarter. Capabilities that absorb the thing you were about to pay for.
Three live examples from the last eighteen months:
- Brokerages that built or bought showing-feedback tooling in 2024 have watched Zillow, MLS portals, and major CRMs absorb that capability for free.
- Brokerages that spent mid-six-figures on custom agent-productivity dashboards are now being undercut by platform-native equivalents bundled into existing contracts.
- Brokerages that rushed to integrate standalone AI writing tools paid license fees for capabilities that are now default features of the email clients their agents already use.
In each case, wait would have been a positive-ROI decision even if the waiting period produced zero capability in the interim. The option value of not being locked in is real, and almost nobody calculates it.
What "wait" is not
Wait is not "ignore it." Wait is an active position. If you're in the wait quadrant, you should be doing three things:
- Tracking the platform roadmap. The roadmap for your CRM, your transaction management system, and the two or three AI platforms relevant to your workflows. Quarterly. Written down.
- Preserving workflow flexibility.Don't let "wait" mean "keep doing it the hard way." Use the waiting period to clean up the workflow, simplify the data model, and make sure whatever ships next can plug in without a migration project.
- Running a cheap tripwire.Pick a light-touch indicator that tells you when "wait" becomes "buy." Usually a specific platform release, a specific cost threshold, or a specific agent-complaint volume.
How this changes the build case
The implication is uncomfortable for anyone who enjoys building: the pure-build case has gotten narrower, not wider, in the AI era. If the capability will be commoditized, building it is destroying value. Building only makes sense for the workflows where you are genuinely differentiated, where the capability is durable, and where the cost of being wrong is bounded by your own discipline to kill it early.
In my operating experience, that's the agent-enablement layer, the compliance policy engine, and occasionally the transaction-orchestration layer. Everything else is either rentable or imminent.
The one-page test
Before any brokerage tech decision (build or buy), write one page that answers three questions:
- What is the half-life of this capability? Name the specific platforms likely to absorb it, and the evidence supporting that timeline.
- How differentiated is the workflow this touches? If a competitor can buy the same vendor tomorrow, it's not differentiated.
- What does being wrong cost? In dollars, in time, in agent trust, in organizational memory.
If you can't answer those three questions in one page, you aren't ready to decide. If the one page ends in "wait," you save seven-figure mistakes. If it ends in "build," you know why. And more importantly, you know when to stop.