What a Steel Trader Taught Me About Selling AI Consulting
I want to talk about the business side for a minute. Not what I built — how I sold it and what I learned.
I Didn't Pitch AI. I Watched Someone Work.
My client wasn't looking for "AI solutions." He was a steel trader doing his thing. I got on a call, asked him to walk me through his day, and I just watched.
Within 30 minutes I saw:
- Data entry that was eating 2+ hours a day
- Information getting lost because he could only focus on one thing at a time
- Decisions being made on gut feeling because historical data didn't exist
- Relationships suffering because he couldn't remember conversation details
He knew all of this. He'd just accepted it as "how the business works."
The Demo That Closed the Deal
The first thing I built wasn't the full system. It was a demo of what was possible.
I scraped one day of auction data, ran it through a basic pricing model, and showed him: "Here are today's items ranked by deal quality. Here's what similar products sold for last month."
His reaction: "Wait, I can see this every day?"
That was the sale. Not a pitch deck. Not a proposal. A working demo that showed him something he couldn't unsee.
What Surprised Me
The aha moments were mutual. Once we could see aggregated sold data by spec family, we both spotted trends that changed how he thinks about his business. I wasn't just a vendor building to spec — I was a partner helping him see his market differently. That's how you get referrals.
Scope expands naturally when you deliver. Started with price monitoring. Then he wanted deal scoring. Then sold tracking. Then CRM. Then call transcription. Then AI summaries. Each one was a natural "what about..." conversation that happened because the previous thing worked well. I didn't upsell. He pulled.
Simple infrastructure = real margins. The whole system runs for ~$50/month. My retainer is 10x that. I spend a few hours a month on maintenance and new features. Multiply across a few clients and this is a real business.
Domain knowledge compounds. I now understand steel trading. Spec families, pricing dynamics, auction mechanics, supplier relationships. The next steel trader I work with gets a system in days, not weeks. And I can speak their language. That's a moat no SaaS builder has.
The Playbook
- Find a small business owner (warm intro, local business, LinkedIn)
- Ask to watch them work for 30 minutes
- Identify the manual workflow they've stopped questioning
- Build a small demo that makes them say "wait, I can have this?"
- Deliver, collect feedback, expand scope naturally
Stop building SaaS for imaginary users. Go sit next to a real person and watch them work. The problems are obvious once you're looking.