Watercolor landscape

From Business Cards to Meaningful Connections: Building My AI Networking Tool

Hite labs profile

"Anyone Can Build an App"

You've probably seen the headlines: "Build apps with AI in minutes!" "No code required!" "Just describe your idea and watch it appear!"

Here's what they don't tell you: Throwing an idea into Claude Code or Lovable will give you something. But that something probably won't work for anyone except you. And it definitely won't solve the actual problem you set out to fix.

The gap between "AI generated an app" and "I built a useful app" is bigger than anyone admits.

So I'm building this networking tool in public—not just to show you the end result, but to walk you through what actually goes into building an app that other people will want to use.

Because here's the truth: AI tools are incredibly powerful accelerators. But they can't replace the thinking that comes before the building.

What This Series Will Cover

Over the next few weeks, I'll be building this networking follow-up tool in vertical slices—one complete feature at a time. Each post will show you the full journey from idea to working demo:

For Each Feature, You'll See:

  • What am I actually solving for?
  • How should it work for real people?
  • What tools did I choose and why?
  • How did I get AI to build what I need?
  • What did I learn when it didn't work the first time?
  • Here's a demo of it actually working

Plus the realistic stuff:

  • What it actually costs to run
  • How I'm keeping user data safe
  • How I'm getting it deployed

This is the 80% that AI tools don't teach you. But it's the 80% that determines whether your app actually solves a problem or just adds to the noise.

No theory-only posts. No "I'll build this eventually." Every post shows you something working.

The Networking Paradox

Since starting Hite Labs, I've been going to a lot more networking events. Chamber mixers, industry conferences, local business meetups. And there's this incredible high when you're in the moment—connecting with people, discovering shared challenges, realizing "oh wow, we could really work together on that."

The conversations are energizing. Someone mentions their invoicing nightmare, and you immediately see how you could solve it. Another person describes their team's process bottlenecks, and you're already mapping out the solution in your head. You exchange cards, promise to follow up, and genuinely mean it.

Then the next day hits.

You're back at your desk, buried in the work you missed while you were out networking. The business cards are in your pocket. Your notebook has scribbled notes that made perfect sense yesterday but are cryptic now. And you think: "I need to follow up with those people. That's how I get business."

But following up properly requires:

  • Manually typing contact info into your CRM (or even just your phone)
  • Deciphering your notes to remember what you discussed
  • Crafting a personalized email that sounds natural, not templated
  • Actually sending it before the connection goes cold

The real cost? 30+ minutes per contact. And when you have five, ten, fifteen new contacts from an event—that's hours of work competing with everything else demanding your attention.

So some follow-ups happen quickly. Some happen eventually. And some... well, some have the best intentions but require just a little too much effort. And they never happen at all.

I was complaining about this to my wife. She was heading to a conference, and I remembered that familiar pain—the guilt of cards sitting in a drawer, connections fading, opportunities slipping away.

And I thought: "We can build an app for this."

The "Aha" Moment

We have all the technology we need to make this friction disappear.

You can:

  1. Take a photo of a business card at the event
  2. Record a quick voice note about what you discussed (30 seconds while it's fresh)
  3. Get a personalized follow-up email generated automatically, ready to review and send

That's it. The whole process happens in under 2 minutes, right there at the event or on your drive home.

No typing. No trying to remember details later. No letting connections go cold while you deal with everything else on your plate.

Why I'm Building This In Public

Four reasons I'm documenting this entire process:

1. Accountability

When I decided I wanted to build in public, I wanted that pressure to actually finish it. Trust me, I have plenty of half-built projects in private repos that never got completed. Announcing it publicly means I have to follow through.

2. Teaching the Full Picture

Most "AI app building" content shows you the sexy part (watch Claude generate code!). I want to show you the unsexy but necessary parts: problem validation, user research, wireframing, error handling, cost modeling, security considerations.

3. Proving the Methodology

I help businesses build custom applications that people actually want to use. This is a live demonstration of that process—from identifying a real problem to shipping a solution that solves it elegantly.

4. Live User Feedback

You reading this article right now? That's live feedback on whether this problem resonates, whether the solution makes sense, whether this is something people actually want. If you're interested in following along, join the email list. Your input shapes what gets built.

You'll see every decision, every trade-off, every "wait, that didn't work" moment.

The Technical Revelation

Here's what makes this possible with modern AI:

AI Vision (OCR on Steroids)

Traditional OCR just reads text from images. You get back a blob of words and you have to figure out what's what.

AI Vision understands context:

  • "This is a name"
  • "This is an email address"
  • "This phone number format means it's mobile, not office"
  • "This is the company name, not a person's name"

Claude can look at a business card photo and extract perfectly structured contact data. It knows that "VP of Operations" is a title, not a name. It knows that the thing at the bottom is probably a website URL even if it doesn't have "www" in front of it.

That distinction is huge. It's the difference between getting back usable data and getting back a mess you still have to clean up manually.

Voice-to-Insight

I hate typing on mobile. Especially longer thoughts about complex business conversations.

But talking? That's natural. That's easy. That's how I actually think.

Voice notes capture the nuance of conversations:

  • "They mentioned struggling with manual invoicing, eating up about 15 hours a week"
  • "Really interested in process optimization, mentioned their team is growing fast"
  • "Looking to make a decision in the next quarter"

This context is gold for follow-up. And capturing it takes 30 seconds of talking instead of 5 minutes of typing.

Smart Email Generation

Here's where it gets interesting:

AI can take:

  • Your voice notes (the conversation context)
  • Your contact info (who you're writing to)
  • Your pre-written email template (your voice, your message, your value proposition)

And create a personalized follow-up that actually sounds like you wrote it.

Not a generic "It was nice to meet you" template. A real email that references what you discussed, positions your solution relative to their specific problem, and moves the conversation forward naturally.

Building in Public: The Approach

I'm not following a traditional waterfall process here. Instead, I'm building in vertical slices—one complete feature at a time, from concept to working demo.

The Core Features

Feature 1: Image OCR - Take a photo of a business card → Get structured contact data

Feature 2: Voice Notes - Record quick context about the conversation → Capture what matters

Feature 3: Email Generation - Combine everything → Get a personalized follow-up ready to send

Feature 4: Deployment & Distribution - Get it working on my phone, then figure out how to make it available to others without blowing out my budget

Why Vertical Slices?

Each post will show you a complete feature—from planning to working demo. Not "here's 3 weeks of wireframes," but "here's how I built image OCR and here it is working on my phone."

This is how real development happens: Iterative, imperfect, learning as you go.

You'll see the decisions, the pivots, the "wait that didn't work" moments. And most importantly—you'll see something working at the end of each post.

The Steps I'm Following (But Not in That Order)

Here's the truth: I don't have a rigid 6-week plan. That's not how real development works, and that's not how I work.

But I do have principles I'm following:

Before I Build Anything

Problem validation: Talked to business owners who network. Asked: "What makes follow-up hard?" Confirmed: This isn't just my problem.

MVP scoping: Listed every feature this could have. Ruthlessly cut to three essentials. Everything else comes later.

As I Build Each Feature

Think through the user experience:

  • What does this feel like to use?
  • Where does it break down?
  • What happens when things go wrong?

Choose the right tools:

  • What's the simplest way to build this?
  • What will this cost to run at scale?
  • Can I actually deploy this?

Test with reality:

  • Real business cards (not mock data)
  • Real phones (not just my laptop)
  • Real networking scenarios (not ideal conditions)

Ship and learn:

  • Get it working
  • See what breaks
  • Iterate based on feedback

The Messy Truth

Some days I'll work on UI. Some days on infrastructure. Some days I'll completely change direction because something didn't work how I expected.

That's the real process. And that's what I'll show you—not a sanitized, perfectly linear progression.

The Bigger Vision (But First, Make It Work)

I have ideas for where this could go. But I'm not building those yet.

What Could Come Later

Batch processing: Process multiple cards at once after an event

CRM integration: Push contacts directly to ZoHo, HubSpot, Salesforce etc.

Event tracking: See which networking events produce actual business

Analytics: Learn which conversations lead to meetings

Team features: Share contacts across your team

The Monetization Question

The Cost Reality: Each contact costs roughly $0.05-0.07 to process (Claude Vision + text generation APIs)

The Options:

  • Free tier: 10 contacts/month
  • Freemium: Basic free, advanced paid
  • Marketing tool: Completely free to drive consulting leads
  • B2B pricing: Charge companies, not individuals

My Current Thinking: Build it for myself first. If it works, figure out monetization based on who else finds value in it.

The real question: Is this a product or a marketing asset? Time will tell.

The Realistic Stuff Nobody Talks About

What Will This Cost to Run?

Per contact processing:

  • Claude Vision API: ~$0.03-0.05
  • Text generation: ~$0.01-0.02
  • Storage: negligible
  • Total: ~$0.05-0.07 per contact

At scale:

  • 100 users × 10 contacts/month = $70/month in API costs
  • Free tier sustainable? Yes, if I cap it
  • Paid tier pricing? Probably $5-10/month for power users

You'll see the actual costs as I build each feature. No estimates—real numbers from real API usage.

How Do I Keep Data Safe?

The sensitive stuff:

  • Business cards = PII (names, emails, phones)
  • Voice notes = business conversations
  • Email templates = competitive info

My approach:

  • Encrypted storage for images
  • Delete voice recordings after transcription
  • Transparent about where data goes
  • User control: delete anytime

The AI question: Claude doesn't train on user data, but users deserve to know their cards go through Claude's API.

How Do I Deploy This?

Starting with web app:

  • Works on any device immediately
  • No app store approval needed
  • Deploy updates instantly
  • Iterate quickly

Probably as PWA (Progressive Web App):

  • Installable like native app
  • Camera access like native app
  • Works offline for photo capture
  • Best of both worlds

You'll see the deployment process when I actually deploy each feature. Not theory—actual URLs you can test.

What Makes This Different from "Just Prompt an AI Tool"?

Anyone could prompt Claude Code: "Build me a networking app that scans business cards and generates follow-up emails."

What you'd get:

  • Something that technically works
  • Probably looks okay
  • Might handle the happy path

What you wouldn't get:

  • Error handling (what if the card is blurry?)
  • Mobile optimization (actually works on your phone at events)
  • Security considerations (how is data stored?)
  • Cost-effective API usage (not burning through credits)
  • User experience design (feels natural to use)
  • Real-world testing (works with actual business cards)

That gap—between "technically works" and "actually useful"—is what this series is about.

The Meta Point: Why This Matters Beyond Networking

This isn't just about business cards. It's about reducing friction between human connection and meaningful action.

The same principles apply to:

  • Client intake processes
  • Sales follow-up systems
  • Event registration workflows
  • Customer feedback collection
  • Any scenario where data collection + personalization = opportunity

The lesson: When you combine AI capabilities (vision, voice, generation) with real user pain points and thoughtful design, you can eliminate entire categories of busywork.

But only if you do the thinking first.

What I'm Learning (And What You'll Learn)

Each post in this series will share:

Technical reality checks:

  • What AI can actually do (vs. what the hype says)
  • When tools work great (and when they fail spectacularly)
  • Cost trade-offs at every decision point
  • Performance issues you don't see coming

Development lessons:

  • Building for mobile-first (harder than it looks)
  • API integration challenges
  • Error handling nobody thinks about
  • Testing with real conditions, not ideal ones

Process insights:

  • When to plan vs. when to just build
  • Knowing when "good enough" is actually good enough
  • Iterating based on what breaks
  • Cutting scope without cutting value

Business thinking:

  • Free vs. freemium vs. paid
  • When features become liabilities
  • Building for yourself vs. building for others
  • Knowing when to ship vs. when to polish

No theory. Just what actually happens when you build something real.

The Bottom Line

This is what I do for clients: Identify operational friction and build solutions that feel effortless to use.

Whether it's a networking follow-up tool, a client onboarding system, or an internal workflow app—the methodology is the same:

  1. Understand the real problem (not just the symptom)
  2. Design for human behavior (not just technical capability)
  3. Build for delight (people actually want to use it)
  4. Test with real users (in real conditions)
  5. Iterate based on feedback (not assumptions)

If this resonates with you—if you have operational bottlenecks eating up your time—this is exactly the kind of problem I solve for businesses.

What's Next in This Series

I'm starting with Feature 1: Image OCR - turning business card photos into structured contact data.

Next post you'll see:

  • The planning: What does "good" look like for business card OCR?
  • The build: How I used Claude Vision API and what surprised me
  • The demo: Video of it working on my phone with real business cards
  • The costs: What it actually costs per card to process
  • The lessons: What worked, what didn't, what I'd do differently

Then we'll tackle voice notes, then email generation. Each feature gets its own deep-dive where you see the complete journey from idea to working code.

No multi-week planning phases. No theory without practice. Just building and learning in public.

Let's Talk About Your Idea

Have an idea you'd like to build but not sure where to start?

I help business owners and entrepreneurs turn operational pain points into custom applications that actually get used. Whether you want to build it yourself with guidance or have me build it for you, let's talk about what's possible.

Schedule a Free Discovery Call

Follow This Build Series

I'm documenting the entire process—from problem validation to deployment. Subscribe to get weekly updates on what I'm learning, the tools I'm using, and the mistakes I'm making (so you don't have to).

Subscribe to Newsletter