Kyle Triplett thinks property managers are grading AI on the wrong test. According to AppFolio’s chief product officer, the measure that matters isn’t how many features a tool ships, but how many jobs it finishes end to end inside the system an operator already runs on.
That thesis drove AppFolio’s Realm-X announcements at its FUTURE 2026 conference in September. The company rolled out tools for building workflows, dashboards, and scheduled analyses from the policies and data already in its Performance Platform. It also introduced a Receptionist Performer that answers and routes inbound calls, expanded its Maintenance Performer, and made its Realm-X Connector for Anthropic’s Claude generally available. AppFolio says AI actions on its platform are up 7x year over year.
The core pitch is a familiar one from a platform vendor: consolidate AI on the system of record instead of stacking point solutions. Insights by Blueprint spoke with Triplett about where that argument holds up and where it gets complicated. Topics included which operational functions are overhyped, whether “fix the data first” is an excuse to wait, when enterprise operators should build their own AI, what metrics a board should hold an AI program to, and what governance needs to be in place before an agent acts on its own. The interview has been edited for length and clarity.
Insights by Blueprint: What separates operators getting real results from AI from those still running pilots that go nowhere?
Kyle Triplett: Operators getting results stopped measuring AI in features and started measuring it by completed jobs. Results come when AI is doing a piece of the work end to end, inside the system the business already runs on.
In property management, operators who broadly adopt AI project 31% portfolio growth, compared with 12% among non-users, according to AppFolio’s 2026 Benchmark Survey. Daily use matters, too: 72% of daily AI users report NOI increases, according to the 2026 IREM-AppFolio AI Adoption Report.
We see that gap in what we call the RPM (Real Estate Performance Management) Maturity Model. Operators in the early stages often use point solutions to do existing tasks a little faster. Operators seeing stronger results connect AI to daily workflows on a unified platform. They use it to build capacity and improve outcomes for residents, teams, owners, and investors.
Which operational functions are most overhyped for AI right now, and which are underrated?
The most overhyped idea is that AI should operate autonomously everywhere right away, or that automation is an on/off switch. It’s a dial. Operators should choose, per workflow, how much runs on its own and how much a person touches, and move that dial as trust builds, on their timeline. In a business as operationally complex as property management, trust doesn’t build that way. AI needs clear guardrails, a well-defined role, and a path for people to step in when judgment matters.
Underrated opportunities include back-office financial work and after-hours moments that often fall outside a team’s capacity. Data shows accounting has the lowest observed AI usage but the highest perceived growth potential for automation like continuous financial close and automated invoice routing. AI can also help teams respond quickly after hours, gather the right information, and route work to the right place.
The goal is not to remove people from the process. It is to give them more capacity by handling straightforward work consistently, while keeping teams in control of the decisions, exceptions, and customer moments that require judgment
Most operators say their data isn’t ready for AI. How clean does it actually need to be before AI is useful, and is “fix the data first” sometimes an excuse to wait?
Data quality matters, especially when AI helps a team make decisions or take action. But “we need to fix all of our data first” can become a reason to wait indefinitely.
The better question is whether the AI has the right, governed information for the work at hand: current operational data, relevant permissions, policies, and workflows. Our research shows only 15% of operators identify data quality as their top AI challenge. More than twice as many, 37%, point to accuracy and reliability. That is often a context problem. Generic tools can produce a plausible answer without understanding the rules or relationships that shape a property-management business.
Your data doesn’t need to be perfect. It needs to be the data the business actually runs on, with the permissions and policies that go with it, so the AI is working from your reality instead of a generic one.
Should an enterprise operator build any AI in-house, or is that a losing game against platform vendors?
It is not an either-or decision. Operators are experimenting, and that’s healthy. There will always be cases where someone wants to build around a genuinely differentiated need. Be clear-eyed about the difference between a prototype and something you can run the business on.
Getting a model to answer is the easy part. The control point is everything around it: live data, permissions, keeping it working when the business changes. That’s where most in-house efforts stall, and it’s why a third of the operators who’ve tried say it wasn’t useful in practice. Build where you have a genuinely different need. Lean on the platform for the part nobody wants to maintain.
What metrics should a board or investment committee actually hold an AI program to: hours saved, headcount, NOI, resident retention?
Start with the outcome the capability was meant to change and pair every operational signal with the business number it rolls up to. In leasing, conversion is the signal; days vacant and occupancy are the outcome. In maintenance, time to resolution is the signal; cost per work order and renewal are the outcome. In accounting, close time is the signal; on-time owner distributions are the outcome.
A board holds the program to a higher level: NOI and resident retention. Hours saved are a leading indicator, not the finish line, and headcount isn’t the goal. The question is what the organization does with the capacity AI creates: better service, more growth, stronger margins. That’s where it shows up in NOI.
Then track the guardrails: accuracy, escalation rates, and adoption. A strong AI program delivers business value without adding risk or complexity.
Do you expect software pricing to move from per-unit to outcome-based or usage-based models as AI does more of the work? What should operators push for in contracts?
I don’t expect outcome-based pricing to take over. Operators want to know what they’ll pay. I do expect a predictable foundation with consumption as a growing element, because when software does the work rather than records it, usage becomes a real cost.
In contracts, push for clarity — what’s included, what triggers additional cost, how usage is measured, what support is provided, and how data, security, and governance requirements are handled. Pricing should be simple enough that an operator can see the line from cost to value — capacity, resident experience, conversion, or growth.
What does good AI governance look like at a real estate company? Who should own it, and what policies should be in place before an AI agent can act on its own?
Good governance isn’t a policy document. It’s an operating discipline shared by business leaders, technology teams, legal and compliance, and the people closest to the workflows being changed. And the industry has room to improve: only 8% of operating real estate companies have a formal written AI policy.
Before an AI capability acts, an operator should be clear on what it’s permitted to do, what data it can access, where human review is required, and who’s accountable. The strongest governance is built into the daily experience — AI works within the customer’s permissions and policies, keeps people in the loop, and makes clear when and how it’s acting.
Do residents actually want to deal with AI, or are they tolerating it? What does the data say about satisfaction when AI handles maintenance or leasing?
Residents do not wake up hoping to interact with AI. They want a fast response, a clear answer, and confidence that their issue will be handled. If AI delivers that, it can improve the experience. If it creates friction or prevents someone from reaching a person when they need one, it does the opposite.
The strongest use cases are where speed and consistency matter. 86% of residents who interact with an AI concierge report satisfaction with property-manager communication. AppFolio’s Maintenance Performer resolves one in four work orders directly with the resident. Leasing Performer has helped customers increase showings by up to 35%, with roughly one-third coming from after-hours leads. Those are moments when residents and prospects need help now, not at the start of the next business day.
The standard should be better service, not automation for its own sake. Good systems handle straightforward moments quickly and make the handoff to a person easy when empathy, judgment, or escalation is needed.
What will AI be able to do in property operations two years from now that it can’t reliably do today?
The short version: AI moves from answering to acting. Two years from now, AI finishes the job instead of helping someone do a task. Today it drafts the renewal offer. Soon, the renewal is handled within the rules the operator set, with a person stepping in only where judgment matters. That’s what will change how operators measure software: not what it can do, but what got done.
But reliability is the bar. A compelling demo isn’t an operating capability. AI has to be accurate, governed, and connected to the workflows where the work actually happens.
What’s one AI decision operators are making now that they’ll regret in three years?
Treating AI as a pile of point solutions instead of a plan for how it fits the operating model.
Point solutions can be useful, and operators should test new capabilities. The risk is adding a tool for every problem without asking where the data lives, how permissions are managed, how work moves between systems, or what happens when an exception occurs. Every new tool adds another data location, another permission model, another exception path. That’s not scale; it’s sprawl.
Start with the business outcome, then build on the platform and workflows your organization already relies on. That’s how operators scale AI responsibly without the unnecessary complexity.
– Nick Pipitone
