Why Multifamily Got the AI Tools and Retail Got the Leftovers

Proptech funding reached $16.7 billion globally in 2025, a 67.9% jump over the prior year and well above the roughly $14 billion placed in 2019. Capital came back.

It did not come back evenly.

There are venture firms whose stated focus is multifamily real estate technology. Reporting on the sector routinely quotes them, and multifamily has its own trade press covering its own proptech cycle. Retail has no equivalent, and the absence is visible in the products.

We should be upfront about a limit here: we could not find a public dataset splitting proptech venture funding by asset class, so we are not going to invent one. What follows is an argument from market structure and from what we can observe directly in the tools themselves.

What we can observe

We maintain a registry of AI tools serving real estate and filter it for retail relevance. Roughly 540 catalogued tools produce 115 that a retail owner-operator might plausibly use, and sweeping directly for the jobs that pool under-covered, CAM reconciliation and retail site selection, added 8 more that no directory had listed.

The exclusions are the interesting part. A large share of what gets cut is multifamily-only by design: resident lifecycle automation, leasing chatbots tuned to apartment inquiries, rent-pricing engines built on unit comparables, renewal management, resident communication platforms. These are not bad products. They are precisely built for a different asset class, and no amount of goodwill makes them work on a strip center.

Meanwhile the work most specific to owning retail, CAM reconciliation, percentage rent, co-tenancy tracking, exclusive use conflicts, is served by enterprise platforms or by nothing.

That is the gap, and it is not subtle once you go looking.

Why it happened

Four structural reasons, none of which involve anyone deciding retail did not matter.

Multifamily is consolidated and retail is not. A single large apartment operator may run tens of thousands of units on one platform. Selling that operator once produces enormous seat count. Retail's ownership is spread across an enormous number of small owners. There are 116,214 shopping centers in the United States, and even the largest owner in the country holds around 15,450 properties, with most well-known REITs far below that. A vendor choosing where to spend its next sales dollar is not choosing between asset classes. It is choosing between one contract and eighty.

The unit of measurement is standard in multifamily and bespoke in retail. An apartment unit is roughly an apartment unit. Rent, term, and renewal follow patterns that support pricing models and benchmarks. A retail lease is negotiated in ways that resist standardization: every CAM pool, exclusive, co-tenancy condition, and gross sales definition is drafted for that deal. Software loves a standard unit. Retail does not offer one.

Multifamily has a data flywheel that retail lacks. Rent-pricing products improve as more operators contribute comparable data. Retail lease terms are confidential, non-standard, and considered competitively sensitive, so the equivalent flywheel never spins.

Residential attracted the consumer-tech talent. Renting an apartment is a consumer experience, which drew founders and investors comfortable with consumer product patterns. Leasing a retail space is a negotiated commercial transaction that requires domain knowledge most software founders do not have and cannot quickly acquire.

What retail actually inherited

Three things, and it is worth naming them plainly.

Horizontal tools that happen to work. Document intelligence, general assistants, and contract review were built for everyone and therefore work for retail. This is the largest genuinely useful category available today, and it arrived by accident rather than by anyone building for you.

Enterprise platforms priced past the tail. Real capability exists for large retail owners. It is priced and structured for portfolios most owners do not have.

A thin layer of purpose-built retail products. Site selection and foot traffic analytics are genuinely retail-native, and both are sold on enterprise terms. Below that, very little.

What retail did not inherit is the middle: affordable, self-serve, purpose-built software for an owner with six centers. That segment exists in multifamily and largely does not exist here.

Whether this changes

Two forces point in opposite directions.

Against: the economics that pushed vendors upmarket have not changed. Fragmented markets are expensive to sell into, and a venture-backed company will keep choosing the bigger contract.

For: the cost of building vertical software has fallen sharply, and general models handle the hardest part of retail's problem, which is reading non-standard documents. The thing that made retail unattractive, that every lease is different, is exactly the thing this generation of technology is good at. A small team can now build something useful for retail without the data flywheel that was previously required.

That is a real change, and it is why we expect the gap to narrow from the bottom up rather than from institutional vendors moving down.

What to do in the meantime

Stop waiting for a retail-specific platform to arrive and solve this. Use the horizontal tools aggressively, because document reading is where your actual pain is and horizontal tools are already good at it. Be skeptical of anything multifamily-native that has been repositioned for commercial, and ask specifically whether it has ever seen a CAM pool or a percentage rent clause.

And treat the absence of purpose-built retail software as information rather than as a verdict on whether AI is useful to you. The tools that exist are not the tools that could exist. They are the tools that got funded.