Is Your Business Idea Actually Good? This Claude Hack Provides the Cold, Hard Truth

Artificial intelligence has made it easier than ever to start a business, with tools like Anthropic’s Claude Code and OpenAI’s Codex enabling anyone to build websites, apps, and SaaS platforms. But not all ideas are created equal, and sometimes what seems like a great idea has already been done before, or is unviable for reasons you hadn’t considered. In the past, confirming the viability of your startup or product idea involved a combination of research, analysis, and asking trusted friends, family, and business associates. But now that you have a virtual smart person in your pocket or on your desktop, AI can help streamline this process. In this article, we’re going to teach you how to set up an AI model (in this case Claude) that will give you the cold, hard truth about your business ideas. To start, you’ll need to create a customized version of Claude that’s highly skeptical and critical of your ideas. AI models have often been found to be sycophantic—more interested in telling you what you want to hear than actually giving it to you straight. We need to give Claude a set of custom instructions to follow so it can avoid this pitfall. To start, I opened up the Claude desktop app (you can download it here), opened Claude Cowork (Anthropic’s tool for knowledge work), navigated to the projects tab, and started a new project named “Business Idea Viability.” Next, on the newly created project page, I selected “set project instructions,” and pasted in the following prompt: “You are a skilled business analyst with decades of experience under your belt. You excel at receiving an idea for a business and comprehensively running research to determine if this idea has already been turned into a business, looking into trends, historic parallels, and data related to the proposed business idea. You are highly critical and skeptical of new ideas and difficult to satisfy, but fair when you encounter a legitimately good idea for a business. You are direct in your communication style and ‘tell it like it is’ with brutal honesty.” Finally, I added two articles from Harvard Business School to the project’s files. Those articles were titled “How to Come Up With an Innovative Business Idea” and “5 Steps to Validate Your Business Idea.” With this, Claude should have a solid understanding of what makes a killer business idea. To start, I wanted to see what kind of feedback Claude would give to an obviously bad idea. Here’s the terrible idea I pitched to the model: “Help me validate this business idea: Leadership coaching for dogs. What if house-training your pooch was just the first step on the road to true doggy-disruption? Using a proprietary mix of AI, performance management tracking, thought leadership, and retired K-9 police dogs, we will teach your pet to not just be a good boy/girl, but a leader in their community and home. Optional add-on classes will help dogs learn some classic dog leadership techniques, like saving kids from wells. Does this seem like a viable business idea?” Despite the ridiculousness of my premise, Claude treated it with deadly seriousness. The AI found that “the cultural environment actually does support some version of what you’re describing.” According to Claude, the dog training market is growing at a nearly 10 percent rate, with the biggest driver being “the trend of treating dogs like family members, even like employees with performance goals.” Having said that, Claude was quick to point out that dogs do not have careers (although to be fair, some dogs kind of do), and that “the entire leadership/thought leadership framing is anthropomorphic nonsense that collapses the moment a customer asks what their dog actually learns in Week 3.” Even the more realistic pieces of my business idea, like AI-powered performance monitoring, had already been hit by other startups, like smart dog collar company Fi. Overall, Claude gave my idea a viability score of 3 out of 10, but said the only reason it wasn’t a 1 is that there’s a genuine market for dog training. All right, so the test worked: I knew for sure that Claude would give it to me straight, so next I wanted to try out a genuine business idea. Here’s my second shot: “An AI-powered employee sentiment and internal manager effectiveness survey solution. Instead of forcing employees to take dozens of surveys, our AI interviewer simply pings you on Slack/Teams/GChat, and engages you in a natural language conversation about your current working situation, relationship with your manager, and outlook of the company’s leadership and vision for the future. We could even set up voice-based interviews to capture more data.” Claude gave this one a 6.5 out of 10, and said that while I had identified a real problem of survey fatigue, I was far from the first person to have this idea. The AI assistant identified InFeedo.ai as the main player in this space; the company provides access to an AI agent named Amber, which proactively chats with employees and generates personalized queries. “They are you,” Claude said, “but years ahead of you.” By this point, I was running out of ideas, so I went back to the pet angle with a new concept: “How about a product that is essentially an AI-powered smart pet sprayer? It would look like a flower vase or cylinder that sits on your table, has a 360 camera with AI vision, and can identify when pets jump on a table and spritz them with a 360 water sprayer.” Claude called this idea “genuinely interesting,” and rated it as a 7.5 out of 10, bordering on an 8. Not only is the pet tech market booming, the AI said, but “the existing competitive landscape here is also notably weak.” The only comparable product was the PetSafe SSSCAT, a motion-activated canister of compressed air that sprays when it detects movement within three feet. Unlike this “dumb, blunt instrument,” as Claude described it, my sprayer would be smart, and capable of telling a dog or cat from a human. Claude’s recommendation for a TikTok-ready elevator pitch? “It follows your cat and sprays only them, not your laptop.” With the makings of a solid idea, I asked Claude to use PowerPoint to create a pitch deck that I can use to help potential business partners and investors understand the idea. And there you have it, a validated business idea courtesy of Claude (please don’t steal it). The examples here may be silly, but hopefully the lesson is not. The next time you have a crazy-great business idea, try hashing it out with an AI model; by instructing it to be honest and direct with you, you can get actionable feedback on demand. BY BEN SHERRY @BENLUCASSHERRY