Risks and Reasoning: AI Can Hurt Our Innovation if We Let It

For high school students, summer is as much of a time of growth as any day in the school year—new hobbies, new opportunities, and new chances to network. For me, amidst my summer of businesses, startups, and incubators, I’ve learned a lot. Yet at the end of the summer and beginning of the school year, I find myself left with one question: is it safe to trust AI in the early stages of starting a business? 

I spent the first half of my summer in sunny San Diego at a startup incubator program for teens called LaunchX. The structure of the program was fairly simple: participants were assigned to a random group of teenagers from all over the world and were tasked with starting a company in just two weeks. I’d shown up with all the AI tooling to ensure our business lasted, specifically in the face of the competitive and AI-enabled startup world. 

Before attending this program, I could not think of a single reason AI would be harmful to new, inexperienced founders. I envisioned it as something of the last piece to our business puzzle—grounding us when we were soaring too close to the sun, doing the grunt work and research when needed, and generating designs and slogans. 

So, after some brainstorming, I took my group’s idea to Claude. I asked for its opinion on our ideas, to which it responded, “This idea is not feasible.” According to Claude, our niche was too broad, our margins were too small, our sourcing was too expensive, and our idea had failed before. Immediately, the idea that I was bragging about to my dad on the phone and writing down in my journal felt doomed. I thought if Claude, the never-wrong ultimate business consultant, said the idea wasn’t possible, then it truly wasn’t possible. I brought this concerning revelation to my co-founders along with my itemized list of how we should pivot. To my surprise, they didn’t seem concerned whatsoever. Instead, they explained that I should calm down. While Claude didn’t understand our idea, they said, real humans believed that we had something. Humans with experience in entrepreneurship, degrees, and life in the real world—all things that AI could never replicate, yet are essential for understanding potential for innovation. 

Candidly, my discovery of Claude’s shortcomings did not come at the same time as the rest of my group’s. Instead, it was when I left the program and talked to my dad nearly two weeks later. My revelation was this: Claude was wrong. And, sadly for me, I was too. To make this realization, I had to understand that the beauty of starting companies is that they are built on ideas, risk, and sometimes blind confidence. Founding companies is an endeavor that is uniquely human—it requires human thinking and human emotions to work. When AI judges an idea, it is purely algorithmic. It’s a hodgepodge of past precedent and current data. A tool trained on just past precedent will both structurally undervalue ideas that don’t have strong prior roots in success (in other words, new ideas) and be unable to envision them in the real world, because it doesn’t live in one.

Claude knows nothing about the feelings that come with having a good idea, picturing it in the world, telling other people, and inspiring them too. Similarly, it knows nothing of conviction, drive, and grit—all things that you need to be successful. And without that knowledge, it ends up hampering those qualities, because it tells people “no” before they begin. 

The tools we have built that we think increase our understanding of the world actually hinder the innate human ability to take risks. Nobody knows what AI would have said about the most successful endeavors of our time: Google, Airbnb, or even OpenAI itself. All of these companies were inherent risks to the founders because they were new, exciting gambles. Google’s founders tried to sell Google for $750,000 in 1999 and couldn’t make the sale. Airbnb’s founders had to sell cereal boxes to fund the company after being passed on by investors. OpenAI, founded in 2015, spent over five years being underfunded in the industry. What makes these ideas so successful is that they were so new, innovative, and risky that even seasoned members of their fields weren’t certain they would work. For entrepreneurs, this sort of overreliance and excessive faith in AI means waiting for an idea that Claude would like instead of taking the risk and going for one that you’re in love with.

I believe that this is most harmful not for the AI-educated or the serial company owners, but for the next generation of entrepreneurs. If this kind of thinking takes over, most of these people will have an innate fear of failing. That might lead them to take a backseat and become overcritical of their own ideas just because they read an AI report that rejected them. This doesn’t only affect new entrepreneurs, though—it affects everyone. If truly revolutionary and special ideas aren’t acted upon while safer bets are, society will see less innovation and slower progress.

The impacts of this way of thinking also go beyond entrepreneurship. Entrepreneurs aren’t the only people taking risks. Scientists and researchers will research less compelling hypotheses. Internal employees will be less likely to bring new ideas to the table. Artists and creatives will innovate less. People will be more inclined to do what already works and less inclined to try something new because it feels safer and, according to AI, has a higher success rate. 

So, what can be done? How do we walk the line between utilizing the most influential and powerful technology in the world to remain a player in competitive markets while also not losing the risk-taking, grit, and human essence of this process? The way to do this is to reframe how we think about AI. Instead of thinking of it as a tool with opinions, we need to start thinking of it as a tool with information. Instead of asking AI if our concept was good, I should have asked instead for numbers and market research, or design tips and mockups. AI can provide information and assistance, but we should let ourselves decide what’s best for us. But, deeper than that, I think this requires a reframing of our minds so that we no longer value AI as much as a human opinion. 

This starts at the beginning. AI proficiency will eventually be taught, and with it should come the understanding of when to ask for its opinion versus when to use it as a tool. AI has invaluable uses in automating processes and helping humans do their jobs better, but it should always be just that: a tool that enables humans to do what they do best: think, feel, take risks, understand, and, in my case, start great companies.