Most multifamily firms have now heard every version of the AI pitch. Many still cannot answer a basic question about their own portfolio without three people and a spreadsheet. An applied-AI multifamily program is the difference between the pitch and the answer. In a conversation on the Multifamily AI Podcast, Revolution RE CEO Elizabeth Braman unpacked what it takes to get there: data standardization, a strategy tied to actual business goals, and a team that is part of the work rather than a recipient of it.
Applied AI is AI pointed at a decision someone in your company already makes. Not a demo, and not a pilot that lives in one region. It is a capability wired into the work: underwriting an acquisition, explaining a variance, deciding which property to visit this week.
The distinction changes what you buy and in what order. A firm chasing AI as a category buys a model and looks for a use. An applied-AI multifamily firm starts from the decision that is slow, expensive or unreliable and works backwards to the data it needs. In apartment investing that path leads to the same place first. The data has to be standardized before anything downstream is trustworthy.
Braman’s central argument is that companies get more out of their data when the data strategy is aligned with the company’s overarching goals. The alternative is a strategy assembled from whatever each department happened to need. A data strategy built department by department produces a warehouse full of unrelated answers. One built from the top produces a portfolio that can be compared to itself. That is where applied-AI multifamily work begins: with the goal, not the model.
In practice, alignment means deciding what the firm is optimizing for before deciding what to measure. An owner focused on growth through acquisition needs different data readiness than one focused on operational margin at a stabilized portfolio. Different granularity, different refresh cadence, different comparability requirements. Both need standardization; they need it in different places first.
Data integration in multifamily is treacherous terrain, and Braman is direct about why. Wrangling various data sources into a single, coherent entity is genuinely hard in an industry where standardization can still feel like a distant dream. Different property management systems describe the same unit differently. Different managers book the same expense to different accounts. Different portfolios use the same word for different things.
An applied-AI multifamily system therefore has to be seamless and scalable. It should stand the test of time rather than work until the next acquisition or manager transition. That is an architectural requirement, not a preference. A pipeline built around one PMS breaks the first time a portfolio adds another.
This is the work of an ETL and standardization layer. It extracts from every system of record in the portfolio, transforms each source into a common data model, and loads the result somewhere every downstream tool can reach. The step-by-step standardization process shows what happens to a raw property export along the way. Once the data is in that shape, the same foundation serves reporting, benchmarking and AI-driven analysis without a separate integration project for each.
Braman highlights the importance of crafting KPIs that truly reflect a company’s performance. The qualifier is doing real work in that sentence. Most multifamily reporting packages contain metrics that are present because they have always been present, not because anyone acts on them.
In applied-AI multifamily reporting, a KPI that reflects performance has a few properties worth insisting on. It is defined once, so occupancy means the same thing in the asset management report and the investor package. It is owned by someone who can change it. It is comparable across the properties in the portfolio, which is only possible if the underlying data is standardized. And it is tied to a decision, so a movement in the number triggers something other than a comment in a meeting.
The sequencing problem is worth naming. Firms often try to define KPIs before standardizing data. They then discover the definitions cannot be computed consistently across the portfolio. Standardization is what makes a KPI a measurement rather than an estimate.
The point Braman returns to that gets the least attention elsewhere is the critical nature of involving the entire team in the data journey. Data quality is produced at the site level by people entering information into a property management system. No amount of downstream cleverness fixes an input problem at the source.
Treating standardization as an IT project therefore misses where the leverage is. On-site teams need to understand why a field matters and what decision it feeds. Regional managers need to see their own data reflected accurately enough to trust it. Finance needs to stop maintaining a private shadow version. When those groups are participants, the data improves continuously. When they are not, the quality of the dataset decays between projects. Applied-AI multifamily adoption is won or lost at that level.
Braman’s observation is that historical data reveals more than just past performance. A standardized history of a portfolio is not an archive. It is the training set and the comparison set for nearly everything an operator wants to know going forward. What did turnover actually cost across the portfolio, and where is it drifting? Which properties recovered fastest after a rate change? Which line item explains the variance that came up in the last asset management call?
Those are questions AI answers well when the underlying data is consistent. It answers them confidently and wrongly when it is not. That is why standardization is the game-changing step rather than the preliminary one. The value of AI in the multifamily industry is gated by data readiness rather than model availability.
The applied-AI multifamily use cases that follow are the ordinary ones. The first is automating the repetitive analysis and document work that currently consumes analyst time, the territory covered in AI automation for multifamily operations. The second is giving an assistant structured access to portfolio data instead of a pile of exports. Revolution RE’s MCP server, currently in beta, exists for that second case: a defined way for an AI tool to query standardized property data directly.
Revolution RE’s own framing of the work is deliberately unromantic: tactical insights and reporting for peak apartment performance, because data drives smart operating decisions and knowledge is power.
Connect with Elizabeth Braman for more on data standardization and applied AI in apartment investing. If the conversation was useful, subscribe to the podcast on your platform of choice and leave a rating and review. It helps others find the show. Pass the episode to someone who would benefit from it.
An applied-AI multifamily program aims AI at a decision the business already makes rather than at a category or a pilot. In apartment investing it starts from a slow or unreliable decision and works backwards to the data that decision needs, which in practice means standardizing portfolio data first. The measure of success is whether a named person makes a better decision, not whether a model was deployed.
In the analysis that standardized historical data makes possible — turnover costs, variance explanation, comparative property performance — and in automating repetitive analytical and document work. Historical data reveals more than past performance; it is the comparison set for decisions going forward. The gating factor is data readiness, not model availability.
Because portfolio data arrives from many systems that describe the same unit, expense, or lease differently, and a model trained or queried on inconsistent inputs produces confident answers that are wrong. Standardizing every source into one common data model is what makes historical data comparable, KPIs computable across properties, and AI output trustworthy. It is the first step of an applied-AI multifamily strategy, not an optional one.
Applied AI starts where the data stops being an obstacle: every system of record extracted, standardized into one model, and available to the people and tools that need it.
See plans and pricing