Conversation - June 23, 2026

Building Trustworthy AI for Real Estate

A conversation with Joe Wilhelmy, CTO of Roof AI, about the future of AI-enabled decision making, explainability, and the challenges of building products people actually trust.

Interview conducted June 18, 2026.

Paul Huber By Paul Huber Co-Founder, HomeSignal Conversation with Joe Wilhelmy CTO, Roof AI

1. Decision systems

Successful AI products can promote more consistent and defensible decisions.

As part of HomeSignal's ongoing research into AI, computer vision, and property condition assessment, I recently spoke with Joe Wilhelmy, CTO of Roof AI, about the future of AI-enabled decision making, explainability, and the challenges of building products people actually trust.

One theme surfaced repeatedly throughout our discussion: successful AI products are rarely intended to replace people; rather, they focus on promoting more consistent and defensible decision-making.

One of the challenges we're exploring at HomeSignal is how AI can help reduce subjectivity in property condition assessments.

When discussing appraisal and inspection workflows, Joe offered an important perspective:

"The opportunity isn't to replace the appraisal. It's to help make part of the process smarter, more repeatable, and more consistent."

"Finding a way that you can employ the new tooling and technologies to come up with a more defensible and accurate assessment... that's where the data are bringing new value."

Joe Wilhelmy

This distinction suggests that the value of these technologies is not the ability to replace professionals but the development of more consistent observations and measurements to support professionals in decision-making. This perspective resonated with me given the emphasis on repeatability and defensibility in appraisal and inspection workflows.

2. Explainability

Trust comes from proving that outputs are measurable and understandable.

Many AI products can produce impressive outputs, yet relatively few become trusted business systems.

When I asked what separates trusted AI systems from those that fail to gain adoption, Joe emphasized the importance of transparency and operational rigor.

"You can present the eval sets, the harnesses that you put around this, the operations lifecycle where you're taking the efficacy of the model and using that to further improve the system."

"There are a number of steps that can at least get you to something that looks like proper explainability through a system that is not deterministic at the end of the day."

Joe Wilhelmy

For HomeSignal, this is one of the most important challenges in front of us. Trust doesn't come from claiming that a model is intelligent. Trust comes from proving that its outputs are measurable, repeatable, and understandable.

3. Positioning

The AI label is not a substitute for solving a real problem.

The public conversation around AI continues to evolve.

Joe cautioned that companies should be careful about assuming consumers automatically trust products simply because they're AI-powered.

"I think you've got to be careful where you apply the two letters AI."

"Customers are carrying baggage around hallucinations, job displacement, and all the things that come along with it."

Joe Wilhelmy

That observation raises an important question for anyone building AI products today: are we selling technology, or are we solving a problem?

4. Adoption

The next challenge may be finding the signal through the noise.

Toward the end of our discussion, I asked Joe what organizations might be underestimating over the next 18 months.

Rather than pointing to model capabilities, he focused on the increasing volume of noise relating to AI and the difficulties that creates.

"I think they're overestimating the impact that AI is bringing to their bottom line."

"There's so much overproduction that it's hard to find the signal through the noise."

Joe Wilhelmy

That comment stood out because it applies not only to content creation, but to software products as well. The challenge ahead may not be building AI systems, but instead demonstrating which systems create measurable value.

5. Closing

The future of AI in real estate is likely about supporting professionals, not replacing them.

The conversation reinforced something that we're thinking about frequently at HomeSignal.

The future of AI in real estate may not be about replacing appraisers, inspectors, or property professionals, but helping them to make more consistent, defensible, and repeatable decisions.

Learn more about what HomeSignal is building, read our conversation on trust and expert workflows, or browse more HomeSignal field notes.