Nearly every technology product sold into multifamily now claims to use AI, making the label close to meaningless. But the more important distinction is between products that use AI to do the same work software already did, and products built around things AI makes possible for the first time.
That distinction makes all the difference when it comes to NOI.
Using AI to automate tasks on-site teams already handle may produce some savings. Redesigning the process around a machine doing the work can create much more leverage. It can also solve problems that were previously too small, fragmented, or labor-intensive to bother with.
Dom Beveridge, founder of 20for20, will moderate a panel at Blueprint on this exact topic. “What Does ‘AI-First’ Have to Do With Multifamily NOI?” features Doug Pearce, EVP of IT at Waterton; Aaron Ross, President of Birge & Held; and Jeremy Voigtmann, Managing Director at Harbor Group, each of whom will walk through a deployment his organization is running. Ahead of the session, Beveridge spoke with Insights by Blueprint about the three problems those deployments address, why the NOI benefit usually shows up in headcount that never got hired, and what it all implies for the moats of the technology sector.
BH: “AI-first” is doing a lot of work in the market right now. What separates it from AI-enabled?
DB: Much of what has shipped in the last two years is an AI facsimile of technology that already existed. The test I apply is whether the AI is doing something that genuinely would not have been possible before AI existed, or whether it is a wrapper on a process software has handled for many years.
The distinction is architectural. Most conventional software is a database, a workflow, and a set of screens. When you AI-enable that architecture, you have already constrained what the AI is allowed to do. The technology will no longer be optimal, and there are benefits you will probably never attain.
Take a category like utility billing. An incumbent platform can bolt AI onto what it already has and make the existing process somewhat better. A product conceived from the outset as an AI brain running that process end to end is a different thing altogether, and only one of the two keeps getting better as the models do.
BH: The first of the three cases on the panel applies AI to capital spending. Where is the NOI in that?
DB: AI can now build an extraordinarily precise model of a physical asset, which means it can bring to an individual repair the same planning precision the industry reserves for a full renovation. The delta between the cheapest and most expensive bids compresses, because the model knows exactly what has to be done to execute the repair. You stop getting the conversation where somebody shows up and announces they need more siding material than they thought.
And you accumulate a real record of every repair on the asset, which surfaces systematic problems over time. It’s helpful to think of two very different perspectives on repairs in multifamily. The property manager wants the leak fixed so the resident stops complaining. The capital allocator wants to know whether the capital going into roofs across the portfolio is being allocated efficiently and whether the right vendors are doing the work. AI makes it practical to serve both from one set of data.
BH: The second case is about contracts. Why has that problem stayed unsolved for so long?
DB: Procurement in a large professional organization has two main parts. Sourcing is making sure you have the right suppliers on the right terms; procurement proper is the tactical business of buying from them. Being good at both is how organizations get expenses under control.
Multifamily has both problems structurally, because properties constantly cycle into and out of portfolios, and when you acquire a property you acquire all of its suppliers as well. You inherit a set of drywall, landscaping, and service contracts, and the question is whether they represent the value you have already negotiated everywhere else in that market. The second problem arrives every month, when an invoice lands in front of a community manager who has to determine whether it reflects the terms of the contract that was signed. Solving both is worth several percentage points of NOI, which is substantial upside on asset value.
Every experienced asset manager knows this, but it’s a hard problem to solve because each contract is too small at the margin to justify chasing down. And the tools built for this before AI all require somebody to upload the contract into the platform. And that’s not going to happen in multifamily, where you’re not asking procurement professionals to do it. You are asking service professionals who context-switch across a dozen activities a day. An AI that goes into the shared drives and email accounts on its own, finds anything that looks like a contract, and compares them against one another hands the asset manager what they have wanted for years: where the terms are unfavorable and where the portfolio is overpaying.
BH: The third case is resident-facing automation, which most operators encounter as a leasing tool. What separates a good deployment from an ordinary one?
DB: So far, most of the industry has adopted this technology as a widget that answers phones and nurtures leads. The more interesting deployments use it to deliver more radical change to the operating model.
Adoption went vertical in 2025 because the industry finally got comfortable with resident-facing AI. Leasing ends when the lease is signed, while renewals, collections, and delinquency happen every month and involve residents rather than prospects. There is a flywheel — once an operator sees it work on one of those, usually delinquency, they want it doing every other type of work that has similar characteristics. I think that is why property admin tasks have seen such incredible growth in adoption of AI automation over the last year or so.
BH: So the NOI benefit lives in the operating model as much as in the software.
DB: I think that’s the right way to look at it. Deploying an AI agent as a widget that handles busywork that nobody liked doing may be sensible, but it isn’t the main NOI story. Perhaps a bit more rent gets collected, and collected a bit earlier. But no leverage has been created in the model.
To unlock the leverage of AI, you have to industrialize the process, which means breaking with the way it has traditionally been done at the property. During the panel we’ll talk about a transformational project that, like most of the successes we’re seeing with AI adoption, started with a first-principles design of how processes should work. That usually entails creating a central team to run processes like delinquency. When you introduce AI into that structure it adds infinite capacity to a specialist team. Processes improve substantially, and organizations get far greater leverage out of the model than the property-based model allows.
The question that gets an organization there is what it wants humans to stop doing altogether, and what automation should be handling that people handle today. Companies are getting better at strikng that balance, and it’s enabling them to find performance improvements and change their staffing models. That’s where the NOI benefit really lies.
BH: What ties the three cases together?
DB: In all three, the upside was always obvious and the economics of effort were the barrier. Every contract was too small to chase, every capital project was too administratively onerous to plan well, every centralized process dependent on headcount that ate the savings. AI collapses the cost of applying real diligence to small, fragmented problems, and that is what makes these categories look different rather than just faster.
BH: Do you see this as a harbinger for how technology categories evolve from here?
DB: I do, though I do not know exactly how it plays out. The problems AI can solve are much bigger than the increments at which software has historically advanced. New categories emerge, and AI has this characteristic of not staying in its lane. The successful models will keep eroding what we conventionally thought of as market spaces, which is exciting or frightening at the same time.
There is a related thread I want to put to the audience, which is the “can we not just get an LLM to do this” question that comes up constantly. People asking it seem unaware that the LLM sits on both sides of the relationship. They are thinking about how to avoid buying software, while the developers are using the same tools to accelerate their roadmaps and are generally better at software development than the buyer. The sophisticated organizations are working out where the boundary between buying and building sits.
BH: Last one. What are you looking forward to at Blueprint this year?
DB: Blueprint has a more grown-up atmosphere than the other industry shows. People come with both an investor hat and a user hat on, the audience skews more senior, and the conversation is more relaxed as a result. Vendors tell me all the time that they have fewer conversations at Blueprint but that the quality is higher. Beyond that, this is an unusually interesting moment in the technology sector. There is a general shakeup going on, and Blueprint is where you get the most insight into how it is going.
