AI Says Buy. Should You? A Sane Guide to Investing in the AI Era
Somewhere right now, an investor is asking a chatbot which stock to buy, getting a confident answer in three seconds, and placing the order. The tools have genuinely changed, you can now summarise an annual report, screen five hundred companies, or generate a plausible investment thesis before your tea cools. What hasn't changed is the part that decides whether you make money. Let's separate the two.
First, honest credit. AI is superb at the drudgery of research. Reading a 300-page annual report for the five things that matter, comparing a decade of margins across competitors, translating management commentary into plain language, flagging that a company's receivables are growing faster than its sales. Work that took an analyst a weekend now takes minutes. At our own desk these tools compress the boring layer of research dramatically, and pretending otherwise would be dishonest.
Now the failures, which are quieter and more expensive. AI models are confident even when they're wrong, they can hallucinate numbers, mix up financial years, or state stale prices as current with perfect fluency. They are trained on the past, which means they compress the consensus, ask for a view and you often get the average of everything already written, which is precisely the view already priced into the stock. Making money requires being right about something the crowd has wrong, and a consensus machine is structurally bad at that. And they carry no accountability: when the confident answer is wrong, the model doesn't take the loss. You do.
There's also a subtler trap, and it's psychological. A tool that answers instantly trains you to decide instantly. But speed was never the scarce resource in investing, patience was. The investor who asks an AI for “five stocks to buy today” has already made the real mistake before the answer arrives, the mistake of believing that the bottleneck was information rather than judgment.
So here's the sane division of labour. Let AI do what junior analysts do: gather, summarise, screen, cross-check, and challenge. Ask it to argue against your favourite holding, that's genuinely useful, it will find the bear case you've been avoiding. Verify every number it gives you against the source, exchange filings, company presentations, because fluency is not accuracy. And keep the final decision, the sizing, and the responsibility with the only entity at the table that actually owns the risk: you.
One more thing, specific to India right now. A wave of AI-branded trading products, algo subscriptions and “machine learning tip” services is being marketed aggressively, often by people with no SEBI registration, using the technology's glamour as a substitute for accountability. The regulator's recent crackdown on unregistered advice applies with full force here. An algorithm with no registration number behind it is just a tip with better fonts.
The risk, plainly: both extremes lose. Ignoring these tools leaves you slower than everyone else at the boring parts. Surrendering to them hands your capital to a confident average. The edge sits in the middle, machines for the work, humans for the judgment.
Research over hype, even when the hype computes. Know what you own, and know why you own it.
Prepared by the ETH Research Desk. Prateek Goel is a SEBI-Registered Research Analyst (Reg. No. INH000026390). This is research and market commentary, not investment advice.