Observations and Lessons from Building AI in 2025

In the era of AI, one year feels more like three with the pace of progress and change. It's meaningful to reflect and take stock amidst all of this change.

Observation: People are adapting quickly.
Lesson: Lead with curiosity over fear.

Exemplary knowledge and reasoning is just a prompt away—by mid-year, 1/3 of US adults had used ChatGPT, which is just one of the chat-based LLMs available. This has led to massive changes in behavior, as well as a broader acceptance of what AI technology can do to simplify our lives. It's a technology so intuitive, no complexity is added to utilize it. The fear of emerging tech lays in headlines around job cuts. But the reality is that most property management companies are doing more work with leaner teams already. Technology is the ideal companion to offload the repetitive, low-level tasks and free up humans to work on the higher-value activities. This should be good news for teams. Their roles may look a little different, but they become increasingly important players.

Observation: Work's getting faster, but not easier.
Lesson: Reshape roles and expectations of team members.

With AI taking on lower-level activity, humans have more bandwidth (and responsibility) for tasks that require higher levels of reasoning, interpersonal skills, and complexity. The result is that while more work can get done faster, the work is not necessarily getting easier! We've had to acknowledge this truth internally in team hiring, training, and mentorship. When AI can take on data entry, analysis, and basic communications... people have to step into more challenging roles. It's a time to uplevel and upskill.

Observation: AI can make shiny things even shinier.
Lesson: Start with one focused problem to solve.

It's easy to think too big when ideas suddenly become faster to implement. Shiny object syndrome mixed with vibe coding led to a lot of interesting experiments, but also a lot of time wasted even at the enterprise. We've heard our share of big ideas—amazing to envision, but still really hard to execute on well.

When work can move much faster, building a great AI product experience requires far more clarity on the goal and purpose. This will change how it is structured, trained, prompted. Get clear on the problem this AI implementation is intended to solve, and try to be disciplined on scope and expectation creep. If you're evaluating providers, make sure they're nailing the one thing, really well, before they start selling you side projects. There's so much low-hanging fruit in property management that a well-executed implementation can do far more than a large program that never got out of pilot phase.

Observation: Operational efficiencies can come from places you may not see.
Lesson: AI helps us benchmark what couldn't be easily quantified before.

We have found that while most of our customers know they can be doing better with answering their phones, they don't actually know how big of a problem it is. That's why we love analyzing phone logs to set an initial benchmark.

Our discovery? Most property management companies miss about 60% of their phone calls. Around 10% of those calls drop off in the phone tree, and the other 70% drop off without ever leaving a voicemail. That means that about 50% of calls are missed with no data about why they're calling, who they are, or if an action needs to be taken. Those are missed leads, prospects, and emergencies getting addressed too late—but it's hard to know exactly what with no data.

In an AI-powered world, we can know what the caller's intent was right away. We're also able to measure sentiment, AI, and team performance. We've seen some customers with over 40% of calls being resolved by AI with no human intervention. On average, it's about 34%. We're looking forward to more nuanced insights in 2026.

2026: The year of speed

In 2025, we were all about creating a new category for voice AI. Property management companies were in the exploratory phase of trying out AI in various parts of their business and finding the successful use cases. In 2026, there are more patterns to pull from and tried-and-true use cases. This is going to enable builders and operators to move faster and with less hesitation. This clarity allows them to spend time and energy on implementation and iteration; what will ultimately drive successful AI adoption. It's going to be a year of moving even faster, but with more intention.

‍

‍