For business owners who saw the ad “a million-baht advisor, free with AI”
Popular posts like to claim “you can have a 3-million-baht business advisor for free with AI” or “10x productivity.” The question worth asking is what is real and what is exaggerated.
Short answer: the underlying technique is real and usable, where you feed AI the context of your business and reuse it, but numbers like “replaces a million-baht advisor” and “10x” do not match the evidence, and there are important limits that mean you cannot trust it blindly.
The real technique: feed AI the context of your business
The core of a “Master Prompt” is having AI answer deep questions about your business (vision, strengths, customers, strategy), then saving those answers as a profile you reference again and again. This is proper context engineering, and AI answers genuinely become more accurate and useful once it knows enough of your context.
Tools that support remembering context (as of June 2026):
- ChatGPT Memory remembers across chats, with both saved memories and reference chat history
- Custom Instructions, rules you set once and apply to every chat
- Projects, where you upload files or a profile once and use it across every chat in the project, which suits organizational work
Worth knowing: paid tiers remember context automatically, free tiers may require re-uploading your profile every time, and calling through the API has no Memory, so you manage context yourself in your own system.
Where AI genuinely helps advisory work (with research measuring the results)
AI can draft marketing plans, set OKRs, propose product ideas, and draft SOPs, and research confirms it raises output:
- Advisory work (the Harvard/BCG experiment): AI users completed about 12.2% more tasks, worked 25.1% faster, and produced over 40% higher quality
- Professional writing: time cut by roughly 40% with better quality
- Call centers: 14% more issues resolved per hour on average (newcomers gained up to 34%, while veterans saw almost no difference)
These numbers are real and significant, but they sit in the range of roughly 10 to 40% on specific tasks, not 10x.
⚠️ Limits you cannot ignore
- “Replaces a million-baht advisor / 10x” does not match the evidence. Studies that measured real results (NBER, METR) point to gains of around 10 to 40% on specific tasks. “10%, not 10x” sums it up best.
- AI can always fabricate information (hallucination). This is a structural limit of LLMs, and wrong answers often “sound credible” enough to be hard to spot. Real cases: Deloitte Australia (2025) submitted a consulting report containing fake references, and Air Canada’s chatbot invented a discount condition that a court held the company liable for.
- AI advice is broad and generic if you do not feed it your specific data. Important work needs a human to check it and grounding with real data (such as RAG).
Summary for decision-makers
Use AI as an “advisory assistant” that genuinely speeds up drafting and helps you think, but not as an advisor you can trust instantly without checking. Feed it good business context, use it for work you can verify, and do not turn advertising numbers into your targets.
updated 16 June 2026