AI Property Concierge

The challenge wasn't just design, it was figuring out why an AI-powered service kept breaking trust with users and what to do about it.

I joined the startup as Product Designer on a three-month contract to stabilize the product and build the foundation for long-term growth.

The first problem was personality. A five-star hotel can't sound like student housing, but Ailean had one voice for everyone. I introduced customizable personas with language guardrails that let clients control tone while keeping responses accurate and professional. This wasn't just about branding, it was about making the AI feel like it belonged in each specific context.

The knowledge base situation was worse than expected. Clients were manually updating massive Excel files with property information, and developers had to sync them twice a day. It was slow, frustrating, and error-prone. I redesigned the entire system to be web-based with automatic updates. This didn't just improve the client experience, it gave Ailean real-time access to accurate information, which directly improved how well it could help users.

Testing new features was another bottleneck. Every iteration required developer time, which slowed everything down. I built n8n simulations that let us validate user flows without engineering support, which cut iteration time significantly. I also used AI coding tools like Claude and Windsurf to prototype and ship features myself. Knowing enough code to implement ideas directly meant fewer handoffs, lower costs, and faster deployment.

One of the more practical additions came from research into tenant support requests. Most issues didn't actually require calling in external help, people just needed guidance. I built a curated YouTube video database so Ailean could troubleshoot common apartment problems by sharing pre-approved tutorials. It reduced unnecessary service calls and helped tenants solve things immediately instead of waiting for someone to show up.

The backend needed work too. Clients had no visibility into how Ailean was performing, which made it hard to trust the system. I designed fully responsive dashboards for both product versions with light and dark modes, introduced statistical tracking, and worked through the logic behind the metrics to make sure the data was reliable. B2B clients don't buy products they can't measure.

The three months were intense. I joined during a critical phase when the company wasn't making decisions based on data or evidence, just instinct. By the time I left, human intervention requests had dropped by half, response accuracy was up 20%, and clients were reporting that the new knowledge system had transformed their workflows. More importantly, the startup had the infrastructure and insights they needed to keep improving the product without me. Sometimes the best thing you can do for a company is leave them better equipped than when you arrived.