A study from MIT Sloan found that large language models can provide financial advice that is comparable to — and in some cases better than — human advisors, but only when users frame their questions with sufficient specificity and context [1].

The research tested AI responses against professional financial advice across a range of scenarios, from retirement planning to tax optimization. When users provided clear parameters — income, goals, risk tolerance, time horizon — the AI's recommendations were often indistinguishable from those of certified planners. When questions were vague, the advice quality dropped sharply [1].

The finding cuts against both the "AI will replace financial advisors" and "AI can never replace financial advisors" narratives. The study suggests the binding constraint is not model capability but user skill at eliciting useful output — a dynamic familiar from software engineering but rarely applied to financial planning [1].

On X, the reaction was practical. Users shared examples of well-prompted vs. poorly-prompted financial questions, turning the study into a mini-tutorial on prompt engineering for finance [2]. The most shared takeaway: "AI financial advice is only as good as the question you ask it."

Tech outlets covered the study as a milestone for AI in regulated industries. Business press focused on the implications for the financial advisory profession. Neither engaged the deeper point: the bottleneck has shifted from model quality to user literacy.