Today was a hands-on day, shifting from planning to executing as I dove into building the MVP (Minimum Viable Product) for the AI-powered real estate system. The morning was spent immersed in design—a world I respect but don’t prefer—while the afternoon focused on refining the chatbot knowledge base. This wasn’t just about functionality; it was about creating a polished system that looks great, works seamlessly, and delivers consistent, reliable information to users.
If you’re just tuning in, I recommend starting from Day 1: Laying the Groundwork for an AI-Powered Real Estate System to see the foundation of this project. And don’t miss Day 5: Tool Evaluation and Refining the Approach for insights into how I finalized the tools being implemented today.
The Designer’s Hat: Text Layout and Visual Consistency
The morning began with designing the text layout for the landing page on Carrd, a task I didn’t expect to consume two hours. Adjusting fonts, colours, and point sizes sounds simple, but it quickly became an intricate dance between aesthetics and functionality.
The Design Process
- Font Selection: I tested multiple fonts to find one that conveys professionalism while remaining approachable and easy to read.
- Color Palette: The branding visuals created in MidJourney helped guide the color scheme, ensuring consistency across the website.
- Point Sizes and Layout: Key content areas were designed to draw the reader’s eye, emphasizing critical information like how selling for cash works and lead submission forms.
Minimalism Meets Practicality
My design philosophy was to keep things minimalistic yet effective—providing just enough information to pique interest while avoiding unnecessary clutter. Every element needed a purpose, whether it was guiding users toward the next action or reinforcing the system’s professionalism.
While I don’t consider myself a designer, this experience reinforced how critical attention to detail is in creating an interface that communicates value to potential leads. It also deepened my appreciation for those who excel in this field.
Refining the Knowledge Base: Voiceflow and AI Reliability
In the afternoon, I swapped my designer hat for my engineer’s cap, focusing on developing the knowledge base within Voiceflow. While AI is a key component of this system, I want to ensure that responses remain accurate, consistent, and aligned with the services being offered. This requires a careful balance between leveraging AI and maintaining control over the output.
Establishing Consistency
The goal today was to ensure the chatbot:
- Provides Reliable Information: Responses are grounded in pre-verified knowledge to avoid discrepancies.
- Reflects Service Offerings: Answers align with the specific processes and services outlined on the website.
Using Claude (Anthropic) and Perplexity, I created detailed responses for common queries like:
- “How are cash offers calculated?”
- “What’s the timeline for selling a house for cash?”
- “What does selling ‘as-is’ mean?”
These responses serve as the backbone of the chatbot’s interactions, ensuring users receive clear, accurate, and actionable information.
Addressing AI Hallucinations
One critical challenge when using AI is managing hallucinations—instances where AI fabricates answers that sound plausible but are entirely inaccurate. While AI tools like ChatGPT and Claude are powerful, they occasionally “confidently” provide false information.
What Are Hallucinations?
Hallucinations occur when an AI generates factually incorrect or entirely fabricated responses. While the results may sound convincing, they risk undermining trust if unchecked.
Example: During early testing, I asked the chatbot, “What’s the average timeline for selling a house for cash?” The response confidently stated:
“Selling a house for cash in Canada typically takes 48 hours, including inspections, legal paperwork, and closing.”
This claim, while compelling, was entirely inaccurate. In reality, timelines vary based on factors like location, property condition, and buyer availability.
Minimizing Hallucinations
To combat this, I’m taking a layered approach:
- Pre-Established Knowledge: Relying on verified information generated with tools like Perplexity and Claude.
- Human Oversight: Manually reviewing and curating responses to ensure they align with real-world expectations and processes.
This process reduces the risk of inaccurate or misleading responses while maintaining the efficiency AI brings to the system.
Key Lessons from Today
- Design Isn’t Just About Looks: Creating a polished, minimalistic landing page layout taught me the importance of balancing aesthetics with functionality. Fonts, colours, and layouts aren’t just decorative—they guide users and build trust.
- AI Needs Guardrails: While AI accelerates content creation, it requires human oversight to ensure outputs are consistent and accurate. Building a strong knowledge base is essential for a chatbot to deliver value and inspire confidence.
- Time Flies in the Details: Tasks like designing layouts and refining chatbot responses can consume more time than expected. Attention to detail is key, but so is managing time effectively to keep the project on track.
What’s Next?
Tomorrow, the focus will shift to integrating the chatbot into the Carrd landing page. This involves:
- Embedding chatbot workflows for lead qualification and FAQs.
- Connecting the knowledge base to Voiceflow to test real-world interactions.
- Testing the functionality of automation workflows and calendar integrations.
Additionally, I’ll begin testing the FCT API for property value estimation to ensure it works seamlessly within the system.
Catch up on earlier updates to see how we’ve reached this stage:
- Day 1: Laying the Groundwork for an AI-Powered Real Estate System
- Day 5: Tool Evaluation and Refining the Approach
Stay tuned as the MVP moves closer to becoming fully operational!
Final Thoughts
Day 6 was a mix of creativity and precision. From perfecting the landing page layout to building a consistent and reliable knowledge base for the chatbot, today’s progress reinforced the importance of balancing technical functionality with user experience.
This project is shaping up to be more than just an AI-powered system—it’s a step toward redefining how real estate investors operate. I’m excited to continue integrating these elements and watching the system come to life.
What are your thoughts on using AI while maintaining control over outputs? Let me know in the comments or reach out directly—I’d love to hear your perspective!
Let’s Connect
If you’ve worked with any of these tools or have insights to share, drop a comment below, reach out on social media or email contact@juliandrouse.com —I’d love to hear your thoughts!
For more details, check out my channel on YouTube. Stay tuned as we continue building the future of real estate investing!
0 Comments