Is the current obsession with AI-powered everything blinding us to what it can’t do? We’re told AI is poised to revolutionize business, to democratize innovation, to empower the solopreneur. And while the hype cycle spins, a quieter, more pragmatic truth is emerging: AI is a powerful tool, but it’s still just a tool. The real story here isn’t the dazzling potential of AI – it’s the enduring, irreplaceable value of human judgment, experience, and frankly, good old-fashioned gut feeling.
Tim Desoto, a 49-year-old founder and CEO based in San Francisco, learned this the hard way. In late 2024, he launched an AI-powered shopping platform, Goodlife, despite having no technical background whatsoever. Desoto’s journey, as he shared with Business Insider’s Agnes Applegate, isn’t a tale of AI conquering all. It’s a story of carefully calibrating where AI excels, and where it falls frustratingly short. He didn’t stumble into success by blindly trusting algorithms; he built a business by understanding their limitations.
Based on the original Business Insider report.
Desoto’s initial approach, like many, was to treat AI as a universal solution. “People might think of AI as a hammer, treating everything else as a nail,” he explained. But he quickly discovered that this “hammer and nail” mentality doesn’t hold up. He relied on his network in San Francisco, attending meetups and developer conferences, to gauge which tools were genuinely delivering value. The current buzz around “agentic workflows” – projects like OpenClaw and Claude Cowork – highlights a shift in focus. It’s no longer just about what AI agents can do, but how to run them reliably and securely at scale. This is a crucial distinction, and one that often gets lost in the breathless coverage of new AI releases.
He’s built a “paid stack” including models like Claude Max, Gemini Ultra, and ChatGPT Business, alongside AI-powered productivity tools like Cursor and Figma Make. He’s impressed by the recent improvements in Gemini’s image generation capabilities – faster performance, more stable reasoning, and stronger multimodal capabilities. But even with these advancements, Desoto employs what he calls an “AI conveyor belt” exercise: feeding the same prompt to multiple models, comparing outputs, and actively seeking dissenting opinions from the AI. He deliberately tries to provoke disagreement, recognizing that many models are prone to being overly agreeable. This isn’t about distrusting the technology; it’s about mitigating its inherent biases and limitations.
The turning point came when Desoto was “vibe coding” the alpha version of his product. He could get the code to roughly 95% accuracy using AI against AI, but the remaining 5-40% errors were a persistent roadblock. He lacked the technical expertise to diagnose and fix the underlying issues. This is where the limitations of AI became painfully clear. It could generate possibilities, but it couldn’t provide the nuanced understanding required for complex problem-solving. He ultimately contracted developers, accelerating his product’s development and scalability in ways he couldn’t achieve alone. As he put it, “As much as I can do with AI, it's amazing what technical people can do with AI tools that a non-technical person can't.”
But even with a skilled development team, Desoto recognized the continued need for human oversight. He formalized an advisory board, seeking guidance from experienced mentors and industry experts. These advisors not only provided valuable insights but also helped him identify potential blind spots and connect with strategic partners. This highlights a critical point: AI can augment human capabilities, but it can’t replace the value of diverse perspectives and seasoned judgment.
Desoto now has a clearer understanding of what he can trust AI for. While issues like hallucinations and agreeability can be mitigated, long-term strategic judgment and, crucially, taste still require human oversight. AI can generate options, but humans must choose the right direction. This isn’t a condemnation of AI; it’s a realistic assessment of its current capabilities. It’s a reminder that technology, no matter how advanced, is ultimately a tool, and its effectiveness depends on the skill and wisdom of the person wielding it.
Looking ahead, expect to see a backlash against the “AI-first” approach. The next 18 months will be defined by a reckoning: businesses will be forced to confront the gap between AI’s promise and its reality. The question won’t be “Can AI do this?” but “Is it worth letting AI do this, given the potential for errors, biases, and the loss of human nuance?” We’ll see a surge in demand for “AI whisperers” – professionals who can bridge the gap between technical capabilities and strategic business objectives. And the companies that prioritize human expertise alongside AI integration will be the ones that truly thrive.
