The 8 Patterns Stalling AI Progress In Multifamily

According to the McKinsey Global Institute, agentic AI will unlock $430 to $550 billion in value across the real estate sector. Multifamily is not capturing it.

80% of organizations have deployed AI across at least one business function. Only 20% report a material financial impact. The industry is stuck in what McKinsey calls pilot purgatory: deploying, seeing marginal gains, never moving to transformation.

I spent several months studying why this happens, examining recurring patterns among owners, operators, and PropTech leaders. Eight specific patterns emerged that are stalling progress before it ever gets started, all related to how organizations think about and deploy AI.

You'll find an overview of these 8 patterns below. If you'd like to dive deeper, we also created an AI Momentum Guide with expanded analysis, real-world context, and the reasoning behind why these patterns emerge so consistently.

The 8 Patterns

Pattern 1: Waiting for Perfect Data

The advice circulating tells multifamily to fix their data house first. Get your core systems talking to each other. Resolve the conflicts. Then invite AI in.

Multifamily has one of the most fragmented data landscapes of any industry. If you need to resolve all that fragmentation before you touch AI, many organizations will wait for years.

The real question is not "do we have perfect data?" The real question is: which workflow are you reimagining with AI, and what data does that specific workflow actually require?

Start small and be purposeful about where you bring AI in. One of the key considerations is where you already have fundamental data in place—or data that can be prepared with limited effort. That helps inform where to begin. As you build wins in that first workflow and see positive outcomes, you can continue to strategically expand AI into other areas.

Trying to solve the entire data problem at once is a common mistake. That is a massive undertaking with no clear endpoint, which often leads to wasted time and money. Instead, focus on the data that impacts the workflows you are redesigning with AI. Only work on data issues that will move the needle where you are starting. Everything else gets addressed as you expand, informed by what you learn along the way.

Pattern 2: Waiting for Perfect Stakeholder Alignment

Most organizations believe they need everyone on board before they begin. In practice, that belief is one of the most expensive places to lose time.

Organizations spend months in alignment meetings. Department heads. Cross-functional task forces. IT. Finance. Operations. Because AI is still largely theoretical to most people in those conversations, alignment never fully materializes. Someone has a concern that nobody can fully address. Something feels risky. The group decides to loop in one more stakeholder before proceeding. Nothing starts.

There is one exception that matters: the person leading the organization must be genuinely open to exploring how AI can reshape their vision for the business. Without that openness, nothing else moves forward.

Beyond that single person, alignment does not come from meetings and planning sessions. It comes from results. When the leasing team experiences a workflow where an agent handles 80 percent of lease renewal follow-ups, they buy in. When the maintenance coordinator sees work orders being routed more intelligently, adoption follows naturally. Real proof moves people faster than consensus ever will.

Pattern 3: Waiting for a Complete Governance Policy

Organizations are being told they need a comprehensive, organization-wide AI governance policy in place before they can begin deploying agentic AI. Trying to write that macro-level policy before anyone in your organization has actually worked with the technology is impossible. You end up with theoretical frameworks that become outdated the moment you encounter reality.

Governance is not just an organization-wide policy. It happens at the micro level, within each specific workflow where AI operates, whether you are building AI workflows or deploying a third-party tool. What governance looks like for lease renewal coordination differs from that for maintenance request triage. What data the agent can access, how that data can be used, and when a human is looped in all change based on the workflow.

Build your AI in a controlled environment first. The governance questions that matter become clear through the building process. Let the learning from those micro-level implementations inform your macro-level governance framework.

Governance policy emerges from the work. It does not precede it.

Pattern 4: Assigning AI to the Technology Team

When a new technology lands on the executive agenda, it goes to the teams that always handle technology deployments: IT, systems integration, and infrastructure.

With AI, that default assumption is one of the most consequential mistakes a company can make. AI is not a technology implementation. It is a business and organizational change initiative that requires reimagining how work gets done. It requires people with serious strategic chops at the helm. It requires someone who understands the business deeply enough to identify the most meaningful business outcomes and which workflows should be reimagined to achieve them.

When technology teams lead the strategy, they naturally optimize for what is technically possible. The result is often a well-engineered solution that accomplishes technical objectives but leaves the core work unchanged. Business transformation requires business leadership, not technical leadership.

Pattern 5: Thinking Too Small About AI's Role

Most organizations are still asking the wrong question: where can we add AI to what we already do?

That question will always yield a small answer because it assumes the existing workflow is worth keeping. You end up with faster versions of workflows that might have been broken to begin with. Multifamily is currently chasing a fraction of the $430 to $550 billion in available value by layering AI onto workflows that were never designed for it in the first place. The gap between what the industry is deploying and what is actually possible is enormous.

AI requires a completely different frame. The question is: what would this workflow look like if it were redesigned from the ground up with AI at the center? That is not a minor shift. It changes who does what, when they do it, and why. It changes where decisions get made. It changes the role of the people who currently do the work. When you start with that question, the answers you arrive at are almost always larger and more strategic than what you would have found by layering AI onto an existing workflow.

Pattern 6: Keeping AI at the Leadership Level

Leadership goes into deep research mode. They learn about AI. They build a plan. And one day, they walk into a room and announce that AI is coming.

The team hears one thing: Are we being replaced? That fear is predictable and entirely avoidable. The moment people feel surprised or excluded from something that affects their work, you lose momentum. The moment they feel threatened, you lose trust. And once trust is gone, adoption becomes a struggle instead of an opportunity.

Transparency from the beginning, at every level of the organization, is the only way to build something worth trusting. People closest to the work need to understand what is happening and why. They need to know that their expertise is still valued. That conversation has to start early. It cannot start on the day that AI goes live.

Pattern 7: Leaving Out the People Who Know the Work Best

The people closest to the work know things no process map can capture. They know where the system breaks down and what workarounds have been in place for years. They know what the workflow looks like in real life versus on paper. They know the edge cases that happen once a quarter and cost the organization real money.

Leaving them out of the workflow design process results in agentic workflows built on assumptions rather than reality. When edge cases become failures and workarounds are never documented, the agents break in ways nobody anticipated.

Worse, these same people are the organization's most important advocates for adoption. They are the ones who will use the agentic workflows every day. They are the ones who will decide if it is worth trusting. Cutting them out does not just produce weaker agentic workflows. It produces one that the people who matter most will never fully get behind.

Pattern 8: Failing to Re-envision the Human Role

Most AI deployments have a detailed plan for the agent. Almost none have an equally detailed plan for the human working alongside it.

When organizations design an AI workflow, the focus is almost always on what the agent will do. The human role is assumed without clarity about what the human actually does now, which decisions remain with them, or what happens when the agent hits a limit. There are no feedback loops telling the agent when it is making a mistake.

A human is supposed to oversee a process they no longer fully understand, and is expected to catch errors from a system they were not involved in building. Their job has changed, but their role has never been clarified, leaving them uncertain and disempowered.

Re-envisioning the human role has to be designed with the same level of intention as the agent itself. What does the human do now? What decisions stay with the human? What happens when the agent fails? How does the human provide feedback? These are the questions that must be addressed and shared with the human before deployment.

Go Deeper

Each of these patterns manifests differently within every organization. The frictions they create are specific to your structure, your people, and your strategy.

We created an AI Momentum Guide that explores each pattern in detail with expanded analysis, real-world context, and the reasoning behind why these patterns emerge so consistently. It is written for owners, operators, and PropTech leaders who are serious about moving past pilot purgatory.