I Sell AI Products. Here's Why I Agree You Shouldn't Use ChatGPT for Financial Advice. | Robert Braxton

WIRED laid out five reasons to be sceptical of AI chatbots for financial decisions. As someone who builds and sells AI business tools, I think they're largely right — and the distinction between helpful AI and dangerous AI matters more than most people realise.

WIRED published a piece this week with a headline that might seem like an odd thing for me to share: *"5 Reasons to Think Twice Before Using ChatGPT for Financial Advice."* I build AI products. Brand Promise AI is one of them. I spend a significant portion of my working life telling business owners that AI can transform their operations, cut their costs, and give them capabilities that were previously reserved for large enterprises. And yet: the WIRED piece is largely correct. Here's my honest take. The five concerns — and whether they hold up** The article flags hallucinations, sycophancy, privacy risks from sharing sensitive data, lack of regulatory accountability, and the risk of demotivating human advisors. Let me take each one seriously. Hallucinations**: Yes, still a real problem in 2026. ChatGPT and every other general-purpose LLM will occasionally generate a plausible-sounding answer that is factually wrong, with complete confidence. NYU professor Srikanth Jagabathula put it well in the WIRED piece: "They don't have a notion of ground truth, or what is true." For financial decisions — tax strategy, investment allocation, superannuation planning — a confident wrong answer can cost you thousands. This concern is valid. Sycophancy**: Also real. A study published in *Science* this year documented that AI tools consistently take the user's side in conversations. When you're looking for honest pushback on a financial plan you're emotionally attached to, a yes-bot is the wrong tool. Your accountant or financial planner is supposed to tell you things you don't want to hear. A general chatbot is optimised to keep you engaged, not to be brutally honest. Privacy**: This one I feel strongly about from a cybersecurity perspective. Uploading your bank statements, credit card history, and income details to a general-purpose AI platform — even with training data opt-outs — is not best practice. Your financial data is highly sensitive. The attack surface you create by loading it into a third-party platform is real. Unless you are using an enterprise-grade, compliant deployment with data isolation guarantees, this is a risk not worth taking. Lack of accountability**: A licensed financial adviser in Australia operates under ASIC regulations. They must act in your best interest. If they don't, there are consequences — fines, licence revocation, civil liability. ChatGPT has none of those guardrails. This asymmetry matters enormously when the stakes are high. Demotivating human advisors**: I'll be honest — this one is the weakest argument in the piece. The research cited is about interpersonal dynamics, not financial outcomes. That said, the underlying point has merit: if you're using AI to second-guess a professional without the knowledge to evaluate the comparison, you may end up with false confidence in the wrong answer. So where does that leave AI business tools?** Here's the distinction that the WIRED article doesn't fully draw, and that I think is critical: There is a massive difference between a general-purpose chatbot answering a financial question cold, and a purpose-built AI system operating within a defined scope with human oversight at key decision points. Brand Promise AI doesn't give your business financial advice. It gives you 12 specialist AI executive teams — a CMO to help you think through marketing strategy, a COO to help structure operational workflows, a legal executive to flag risk considerations before you escalate to a lawyer, a growth strategist to help you frame expansion decisions. Every output is a starting point for a human decision, not a replacement for one. That's the design principle that separates responsible AI deployment from reckless AI use. AI is extraordinarily powerful for generating structured thinking, drafting frameworks, identifying considerations you might have missed, and processing information at scale. It is not a licensed professional, and it should never be positioned as one. What I recommend to every client** 1. **Use AI for preparation, not for decisions.** Ask ChatGPT to help you prepare questions for your accountant, not to replace the meeting. 2. **Never upload sensitive financial data to a general chatbot.** Use your bank's own tools, or an enterprise AI platform with explicit data sovereignty and compliance guarantees. 3. **Treat AI output as a first draft.** Always have a qualified human review anything with real financial, legal, or compliance implications. 4. **Understand what your AI tool is optimised for.** A general chatbot is optimised for engagement. A purpose-built business AI tool should be optimised for the specific outcome you need — and should be honest about its limitations. 5. **Keep a human expert in the loop for final decisions.** This isn't a criticism of AI — it's how AI is supposed to work. The best AI deployments I've seen amplify human expertise rather than attempt to replace it. The WIRED article is a useful corrective to the hype. AI is not a magic oracle. It is a powerful tool that, like all powerful tools, can cause significant harm when used incorrectly. That's true of a chainsaw and it's true of a language model. Use them carefully, with the right protective gear, and for the jobs they're actually designed for. Robert Braxton is CEO of Simple Complex Solutions and creator of Brand Promise AI — a structured AI executive department for Australian SMEs. For a conversation about responsible AI implementation in your business, visit robertbraxton.com or call +61 417 226 549.*