Auditing AI-Assisted onboarding steps
One of the most exciting aspects of my work on Dyme was auditing and refining the AI-assisted onboarding experience. This process was designed to ensure that real estate agents could quickly and seamlessly begin building their websites, feeling confident and supported from the get-go.
Having spent five years as a Web Designer specifically focused on real estate agent sites, I brought firsthand knowledge of what makes an agent site not just look good, but work well. I approached the onboarding flow through a practical, design-informed lens—manually building polished real estate websites with specific goals in mind, then reverse-engineering the types of questions and content inputs needed to achieve similar results. This allowed me to spot gaps, refine language, and tighten the logic of the flow, ensuring each step led users toward a website that felt tailored, professional, and truly usable.

To build a more intelligent and personalized onboarding experience, we developed an AI-powered onboarding agent that could ask the right questions up front and translate those answers into a tailored, ready-to-go website. A key part of this work involved defining a range of realistic user personas: from new agents needing a simple digital presence to seasoned teams focused on lead generation or showcasing luxury listings.
Each persona came with distinct business goals, brand personalities, and content needs. By mapping these variations, we were able to determine the minimum set of questions the AI agent needed to ask—while still collecting enough context to generate a website that felt aligned with the agent’s tone, market, and objectives. This approach kept the experience concise, but packed with high-value inputs, making it feel both intelligent and effortless for the user.
Once the core questions and user personas were established, we entered an iterative phase: continuously testing how well the onboarding agent could interpret responses and apply them to the website creation process. We conducted hands-on testing across a range of user scenarios, identifying where inputs were too vague, too complex, or missing key context.
Based on those findings, the engineering team fine-tuned the AI's logic, adjusting how it parsed responses, mapped tone to layout choices, and applied preferences such as colour palettes or lead capture goals. My focus was on ensuring the user experience remained simple and consistent, even as the logic behind the scenes became more nuanced.
The result was a fully functional onboarding bot that asked a small but impactful set of questions, and by the end, delivered a ready-to-go, personalized real estate website, reflecting the agent’s brand, tone, and business priorities.