I'm at a critical decision point and would love feedback from anyone who's built lead qualification systems before. CONTEXT: I'm building an AI automation system for med spa lead generation. Before I start selling anything, I want to validate my assumptions by doing ICP (Ideal Customer Profile) discovery calls with 5-7 med spa owners. But here's the thing: I don't want to waste time calling the wrong businesses. So I'm building a multi-layer qualification workflow to identify which med spas are worth talking to for ICP research. THE WORKFLOW I'M PLANNING: Layer 1: Google Maps Data Scraping • Scrape 200-500 med spas from Austin (using Google Maps API) • Collect: Business name, website, review count, rating, primary category, phone • Use GPT to score leads 0-100 based on these signals • Goal: Identify businesses that are "serious but inconsistent" (my target ICP) Layer 2: Website Analysis • For businesses with websites: Scrape content using Firecrawl • Use GPT to score website quality, service clarity, and alignment with GBP • Combine Google Maps score + Website score → Final ICP tier (A / B+ / B / C) • Goal: Understand which businesses need help vs. already optimized Layer 3: Email Extraction & Outreach • Use Hunter.io to find business owner emails • Send non-sales outreach via Instantly.ai (just asking for 15-min research call) • Track opens, clicks, replies • Goal: Book 5-7 ICP discovery calls (not selling, just learning) Layer 4: AI Voice Follow-Up • If no email response after 5 days: Vapi AI calls them • Script: "I'm doing market research on med spa marketing, would you have 15 minutes to chat?" • Goal: Increase response rate with multi-touch approach Layer 5: ICP Discovery Calls • Conduct 5-7 manual calls with qualified leads • Ask: What's your biggest challenge? What are you spending on marketing? What would you pay to solve X? • Goal: Validate assumptions, understand real pain points, build offer based on

We’ve been building OXB Studio, a two-sided platform designed to help service providers (AI devs, automation builders, technical freelancers) present their work clearly, and help business owners find the right builder without sorting through noise. It’s not a marketplace clone. The focus is on clarity, trust, and execution rather than volume. We just reached a milestone where the core system is working end-to-end, so I wanted to share a few practical build learnings that might be useful if you’re shipping AI or platform products: - Role clarity beats feature depth. The biggest early win was forcing a clear split at onboarding: provider vs client. Most confusion disappears once users immediately know “who they are” in the system. - Dataflow > AI magic. The product only feels intelligent when users can see what’s happening next. Clear states, ownership, and progress mattered more than advanced AI features early on. - Workrooms reduce chaos. A shared space with milestones and scope visibility prevents most breakdowns once real projects start. - Trust signals are everything. Profiles only work when proof (projects, structure, reviews) is surfaced cleanly and consistently. I’m attaching a short walkthrough clip of the current beta so you can see how these ideas were implemented in practice. If you’re building a two-sided platform or AI product, curious what part has been hardest for you: onboarding clarity, trust, or operational flow?

Open any AI tool. It doesn't matter which one! Type this prompt: “Ask me one question at a time to understand what I do, and when you have enough information, clearly explain how you can help me.”

We just had the privilege of letting our partners over at Feeding America know that we will have another 2,000,000 meals ready to send their way in November. We just donated 1,400,000 meals in October and they just informed us that in the last few years we are now at 14,173,890 MILLION MEALS DONATED! It wouldn't be possible without all of you! 😃