“AI lets firms build and stop buying”: Why This POV is Shortsighted

AI Collapses the Cost of the First Build—Not the Data, Compliance, and Audit Layers

For almost two decades, since the PropTech Boom, the smart answer to almost any real estate technology question was: buy, don’t build. The reasoning held. Building meant using engineers you did not employ, leveraging infrastructure you did not run, and navigating a timeline measured in years you could not spare. The PropTech industry existed because that math was honest.

However, the math has changed. A senior analyst with a properly configured reasoning model, a small library of your firm’s own logic, and a free afternoon can now produce working tooling that a venture-backed startup would have needed eighteen months and a seed round to ship. Antony Slumbers made this case recently in a piece worth reading, and he is right about the direction. The buy-versus-build default has inverted for a large slice of what PropTech used to sell. The middle layer, meaning the workflow accelerators, the reporting overlays, and the long list of products pitched as “AI for [some named task],” is now something a capable firm can stand up for itself.

So far, so widely understood. Every operator I talk to has seen the demo, or built one. The part almost nobody has worked out is what happens after the demo. That is the entire question, and it splits your real estate firm’s choice into three paths, not two.

Three (Not Two) Paths to Proptech 

The old framing was binary: buy a product, or build it internally. The inversion adds a third option that most of the commentary skips right past. You can build a solution with a partner who is expert in both AI and real estate, where you own the result and they carry the parts that decide whether it survives contact with your auditors, your regulators, and your own staff turnover.

Here is how the three actually compare once you get past the pitch.

The middle column is where most firms are heading right now, and it is the one worth being honest about.

What you build alone is a toy

This isn’t an insult to your analysts. They are talented, and the tools they spin up are genuinely impressive for an afternoon’s work. The problems are structural, and they show up the moment the prototype leaves the desk it was built on:

• The prototype only works well on clean data—something that real estate does not have. It has a lease that is a scanned PDF from 2009, a chart of accounts that drifts across three property managers, a rent roll that disagrees with the T12, and the same tenant spelled four ways across seven systems. Pour an agent over that and you get a faster way to produce answers that look confident and are wrong, with no source link to catch the error and no one accountable when it lands in an IC memo.
• The prototype only works for one person. The analyst who built it is the only one who knows why it does what it does. She leaves, and you are left with a tool nobody can maintain and nobody can explain. Slumbers waves this risk away as something firms will “engineer out.” Engineering it out is a multi-year knowledge management program closer to what investment banks built for their quant libraries than to anything most asset managers run today. Naming the risk is easy. Closing it is the work.
• The prototype has no answer to the one question that ends every serious conversation about pulling work in-house: Who signs? When a tool informs a valuation, a draw, a covenant call, or an allocation decision, someone is accountable for that number. A demo-quality build gives you nothing to point to and no environment on which to run a regulated workflow.

AI collapsed the cost of the first build. It did not collapse the data foundation, the governance, the audit chain, the drift detection, or the discipline of treating your tooling as institutional memory. That work is ninety-five percent of the cost and a hundred percent of whether the solution survives. A solo build prices the cheap five percent and assumes the rest away.

What an expert partner builds is software with real outcomes

This is the path we built the First Line Software Real Estate Practice around, and the inversion only makes it more correct. 

Here’s how we approach solutions with our clients. 

1. Audit and keep your competitive edge: underwriting logic, the reporting standard, the covenant rules, the agent patterns that encode how your firm reads a market. This is what you own outright: the data products, the integrations, the patterns, the IP. You are not renting your competitive advantage back from a vendor on an annual license, and you are not exposed to lock-in. That is the part of “build, don’t buy” that is exactly right.
2. Add the part the solo build cannot reach: the data foundation. We start where the prototype skips, because nothing above it works until the data is right. Our Unstructured Data Platform (UDP) reads the lease, the rent roll, the T12, the loan package, applies real estate context rather than generic document extraction, and produces clean, source-linked, auditable data your workflows can stand on. Every value traces back to its page in the original document. That is the layer your auditor asks about first. 
3. Focus on a single workflow and expand across your portfolio as wins stack. Make sure the workflows are scoped tightly and proven against your own data. Here are the 9 AI workflows already in production in CRE. 
4. Using our MAIS® framework, we manage your operations with real service commitments, drift monitoring, evaluation infrastructure, a governed skills library that lives as institutional memory instead of one analyst’s private project, and an audit chain that holds when someone has to sign.

The difference shows up in outcomes you can measure, which is the real test of expert software. Underwriting data preparation down forty to sixty percent. Lease abstraction eighty to ninety percent faster. IC-ready files two to three days sooner. Hundreds of analyst hours freed per month and redirected from data entry to judgment. Those are production numbers, on your data, inside your controls, owned by you. A prototype does not produce them, and a generic product does not produce them on the workflows that actually define your firm.

There is one more thing this path gives you that the other two cannot. The “talk to your portfolio” capability everyone wants on day one becomes real once the foundation underneath it is sound. It is an earned capability that arrives after the wins stack, and a firm that tries to start there builds the productivity theater it was trying to avoid.

The window is open now

Slumbers closes with a prediction I agree with. Over the next five years, value migrates out of venture-backed PropTech and back into the operator firms that now have the means to build for themselves. The firms that move first will spend this window compounding capability the rest of the industry will eventually have to procure from someone.

Building it and operating it are two different commitments, and the second one is where a firm either builds durable advantage or quietly accumulates governance debt that surfaces later as a restatement, a failed audit, or a tool nobody can maintain. 

The firms that win will own what is uniquely theirs and bring in an expert partner for the data foundation, the governance, and the operate layer, the same way they already use audit, legal, and quantity surveying for the work they choose not to staff.

For nearly two decades, the smart firms bought what they could not build. For the next five years, the smart firms will build what others would have sold them. The ones that pull ahead will know the difference between a tool that demos well on a Tuesday and software that runs your firm on the days that count.

That difference is the whole game, and it is the one worth choosing a partner for.

Tony McGibbon leads the Real Estate Practice at First Line Software, where the team builds client-owned AI workflow automation for commercial real estate firms across the US and UK. Contact him at tony.mcgibbon@firstlinesoftware.com.