I read something remarkable this week. Michael Wall, who runs a one-person music composition and licensing business, published a detailed account of his year using ChatGPT Pro as what he calls a "first hire." He paid $200/month for OpenAI's Pro tier and documented everything. The results challenge most assumptions about where AI fits in a business.
Wall handles 100-200 commissions annually, releases 5-10 albums per year, and manages a catalog of over 500 tracks. He did all of this while caring for family members. His company's expense ratio dropped from roughly one-third of revenue to somewhere between 3% and 5%. That translates to a 95-97% profit margin.
Those numbers got my attention.
The difference between using AI and working with AI
What strikes me most about Wall's approach is the deliberate framing. He treats ChatGPT as a colleague, not a tool. Every day starts with a new Pro chat that runs until evening. He speaks or types whatever he's thinking about, including business problems, creative questions, experiments that worked or failed, and feelings about particular decisions. At day's end, the model summarizes everything, and that summary becomes the first prompt of the next day.
This pattern matters. Most people I talk to use AI reactively. They have a task, they ask the AI, they get an answer, they move on. Wall built a relationship with context that compounds over time. The model knows his business, his constraints, his goals. Each interaction builds on previous ones.
I've seen this pattern work in my own practice. The companies getting real value from AI aren't the ones with the most sophisticated prompts. They're the ones who've figured out how to maintain context across interactions and treat AI as a persistent thinking partner rather than a search engine with attitude.
From cobbled SaaS to custom infrastructure
Wall's technical journey deserves attention. Before this year, his business ran on what he calls a "Frankenstein collection of SaaS products." He had Shopify connected to various services, a CDN built from AWS components, everything bolted onto Webflow. It worked, but it bled money.
With no coding background, he used reasoning models to build a production Nuxt codebase connected to Stripe, Supabase, and Vercel. Three weeks and four complete rebuilds later, he launched. That same codebase has run for nearly a year with zero downtime. Thousands of commits later, he owns his infrastructure instead of renting it.
This is the part that makes business executives nervous. The idea that someone with no programming experience can build production software feels destabilizing. But the reality is more nuanced. Wall didn't become a software engineer overnight. He built something specific to his needs, with AI handling the technical translation while he provided the business logic and requirements.
The real insight here is about dependency. Every SaaS subscription is a bet that someone else's priorities will align with yours indefinitely. When you can build custom solutions, you reduce that dependency. The cost shifts from monthly fees to upfront learning investment.
Voice mode as a thinking tool
Wall describes voice mode as his "ChatGPT moment." He uses it constantly, asking questions while walking dogs, mowing the lawn, cooking, or doing caregiving tasks. He listens to Deep Research outputs through ElevenLabs Reader, sometimes for 45-90 minutes at a stretch.
This resonates with how knowledge workers actually process information. Reading is fast but shallow. Speaking and listening engage different cognitive processes. When Wall listens to a Deep Research output during a walk and then reopens that chat for a voice conversation about what he just heard, he's doing something sophisticated. He's layering understanding through multiple modalities.
The practical tip he shares is worth noting: add instructions to your custom settings that prevent the model from adding follow-up sentences or offers of assistance at the end of responses. That small change makes voice interactions feel more like actual conversations and less like talking to a customer service bot.
The scheduled learning system
Wall built a collection of scheduled tasks that function like personalized curriculum. Weekly lessons in math, machine learning, design, and market analysis arrive automatically. One schedule delivers detailed breakdowns of different music licensing companies each week, analyzing their websites, pricing pages, terms, and business models.
This approach inverts the typical relationship with information. Instead of searching for what you think you need, you design systems that surface what you should know. The AI becomes a research assistant that works while you sleep.
I find this particularly relevant for professionals trying to stay current on AI developments. The field moves too fast for traditional approaches to professional development. Scheduled deep research on specific topics, delivered and synthesized automatically, creates a different relationship with ongoing education.
What this means for the rest of us
Wall is explicit about privilege. Not everyone can spend $2,400 annually on AI tools. But he also makes a compelling case for ROI calculation. Two hours daily of development work at $50-100/hour would cost $2,800-5,600 monthly through traditional channels. The Pro subscription covers that and more.
The deeper lesson isn't about ChatGPT Pro specifically. It's about approach. Wall succeeded because he:
- Committed to daily practice with the tools
- Built systems for maintaining context across sessions
- Integrated AI into existing workflows rather than treating it as separate
- Focused on his actual business problems rather than chasing features
- Stayed patient through the learning curve
That last point matters. Wall mentions asking questions that "experienced developers might laugh at." The model answered them straightforwardly, thousands of times, without judgment. For people who feel intimidated by technical learning, this patience is genuinely valuable.
The question I keep asking clients
When I work on AI adoption, I ask a simple question: what would change if you had an infinitely patient expert available around the clock who knew your business context?
Most people answer in terms of specific tasks. They'd get help with emails, reports, analysis. That's fine, but it's small thinking.
Wall's example points to something larger. With the right approach, AI can handle the entire support structure of a business, freeing the human to focus on what only humans can do. In his case, that's composing music. The research, planning, infrastructure, and reflection happen in partnership with AI. The creative work remains entirely his.
That division of labor feels like a preview of where knowledge work is heading. The people who figure it out early will have significant advantages. Not because they'll work less, but because they'll work differently, focusing their human attention on the parts of their work that benefit most from human judgment.
The uncomfortable truth
Wall's story also contains an uncomfortable truth for anyone selling AI skepticism. A solo business owner with no coding background built production software, trained machine learning models on his own data, and restructured his entire business infrastructure in a year.
The reasonable response isn't to dismiss this as an outlier. It's to ask what becomes possible when these capabilities spread further. If one person with motivation and $200/month can do this, what happens when the tools get cheaper and easier?
I don't think AI replaces human judgment or creativity. Wall's account supports that view. The music he creates, the decisions he makes about his business, the relationships with his clients, all of that remains human. But the scaffolding around that human work has changed completely.
For executives evaluating AI adoption, the question isn't whether to engage with these tools. It's how quickly you can figure out what Wall figured out: that AI works best when you treat it as a colleague rather than a tool, when you invest in context and continuity, and when you stay focused on your actual business problems rather than getting distracted by features.
The technology will keep improving. The companies that master the working relationship first will have advantages that compound over time. Wall's year-long experiment shows what that looks like in practice.
Full article from Michael Wall : https://www.soundformovement.com/chatgpt-pro-as-first-hire