A Former Goldman Sachs Banker on Using Claude and ChatGPT for Investing — Prompt Design and Financial Literacy Decide the Outcome

The video opens by introducing 'Ochan AI,' a personal assistant built by a former foreign investment banker, populated with all his data. It's been experimentally deployed for participants in his Money Training service, currently at 'version 0.0.' Future versions will follow the Digimon evolution metaphor, the host jokes.

From there the conversation turns. Claude and ChatGPT are tools that visibly split users into 'can use them' and 'can't use them.' Subscribing to a LINE-based conversational AI doesn't help — the 'I don't know what to ask' problem stops most users where they start. The host pays for both Claude and ChatGPT and uses them in real investment decisions. When the interviewer asks 'AI lies, and I don't know what to ask' — a common complaint — the host's reply is direct: 'You're just bad at raising it.'

The host's example of a useful prompt is crisp: 'You are a macro analyst at a foreign investment bank. From the Trump administration's actions and the war landscape, extract this specific viewpoint and explain it using this kind of chart.' Role (macro analyst at a foreign IB), viewpoint (cutting macro through the political situation), output format (explanation with charts) — all specified.

The ability to write at this level of granularity is the dividing line. And the host's blunt observation is that writing well at this level requires financial knowledge on the user's side. AI assists the substance of the analysis — it does not replace the judgment of what to ask. In an investment context particularly, an investor without financial literacy trying to use AI for decisions can't evaluate output quality, which inflates the risk of misplaced confidence.

The second half walks through concrete use cases. AI as a sounding board is the most basic — news collection, point organization, logic checking — but it only works when the instructions are specific enough.

In actual decision-making, the workflow is 'collect information broadly, then have AI organize from this specific angle.' Paying for the subscription means you've committed to extracting more than the subscription cost's worth from the tool. The through-line of the whole video is consistent: AI itself is powerful, but the user's literacy decides the value. For the retail investor audience that the Money Training service targets, the prerequisite of acquiring financial knowledge sits explicitly upstream of integrating AI into investment work.