The problem with being right You have been running your business for a while. You know your customers. You know your market. You know what works. Except sometimes you do not. And the longer you operate on an assumption that happens to be wrong, the more expensive it gets to fix. I have sat with people who were absolutely certain their pricing was right, their target market was clear, their product-market fit was solid. Then we ran the numbers. Or talked to their actual customers. Or just asked a few pointed questions. The certainty dissolved pretty quickly. The issue is not stupidity. It is proximity. When you are inside the thing every day, you stop seeing it clearly. Your assumptions become invisible to you. They just feel like facts. What AI is actually good for here I have built AI-powered tools. I know where the gap sits between what gets promised in demos and what happens in practice. So let me be specific about what these tools can and cannot do. They cannot tell you whether your business model is sound. They cannot predict whether your new product will sell. They cannot replace actual market research or real conversations with customers. What they can do is argue back. If you give a well-prompted AI your core assumptions and ask it to challenge them, you get something useful: a sparring partner that does not care about your feelings, does not have skin in the game, and will not nod along to keep the peace. I have started using this with clients. We take something they believe to be true about their business, feed it into a structured prompt, and ask for counterarguments. Not predictions. Not advice. Just: here are the ways this assumption might be wrong. 73% of the time, according to my notes from the last six months, at least one of those counterarguments lands. The person goes quiet. Then they say something like, "Actually, I had not thought about that." How to do it without wasting time Most people who try this get rubbish results. They type in something vague like "Is my business idea good?" and get vague answers back. The tool is only as useful as the input you give it. Here is what works. First, write down your assumption in one sentence. Not your whole business plan. One specific belief. "My customers choose us because of our speed." "Price is the main barrier to new sign-ups." "Our competitors are not a real threat because they do not offer X." Then ask the AI to generate three to five arguments against that assumption. Ask it to cite the kind of evidence that would support each counterargument. You are not asking for the evidence itself. You are asking what evidence would look like if you were wrong. Finally, look at what comes back and ask yourself: could I disprove any of these? If you cannot, that is a signal. Not proof you are wrong. But a signal that you should probably check. Where this falls apart I have seen people use AI to confirm what they already believe. They prompt it to agree with them, then feel validated. That is worse than useless. I have also seen people treat AI-generated challenges as gospel. They panic because a language model suggested their pricing might be too high, without checking whether the reasoning actually applies to their specific situation. The tool is a prompt for your own thinking. It is not a replacement for it. And it cannot account for context it does not have. If you feed it a one-sentence assumption without background, it will generate generic challenges. You need to give it enough detail to push back meaningfully. That means being honest about your situation in the prompt, which some people find uncomfortable. Why this matters right now The economic picture keeps shifting. Costs are rising. Supply chains are wobbly. There is talk this week about how the situation in the Middle East might push up borrowing and squeeze margins for retailers. Sainsbury's mentioned it. WH Smith mentioned it. These are big companies with teams of analysts, and even they are flagging uncertainty. If you are running a smaller operation, you do not have that team. You have yourself, maybe a few people around you, and whatever tools you can get your hands on. Using AI to stress-test your assumptions is not about being clever with technology. It is about building a habit of questioning yourself before the market does it for you. I have seen people lose months chasing a product extension that their customers never wanted. I have seen people undercharge for years because they assumed their market was price-sensitive, when actually the real barrier was trust. These are expensive mistakes. And most of them start with an assumption that nobody bothered to challenge. What this looks like in practice At ALIRA., we have built tools that structure this kind of thinking. The SWOT Analysis tool at alira.london, for instance, prompts you to articulate threats you might be ignoring. The 5 Whys tool forces you to trace a problem back to its root, which often turns out to be an assumption you did not know you were making. But you do not need our tools to start. You need a clear assumption, a willingness to be wrong, and ten minutes with any decent AI assistant. The point is not to outsource your judgment. It is to sharpen it. What to do this week Pick one assumption you have about your business. Something you have not questioned in at least six months. Write it down in one sentence. Open an AI tool and ask it to generate five arguments against that assumption. Be specific in your prompt. Give it context. Look at what comes back. If any of the counterarguments feel uncomfortable, that is where you should dig. Find one piece of real evidence, from your own data or a conversation with a customer, that either supports or refutes the challenge. That is the work. Not the AI output. What you do with it.