I tried several times builting my own consultant's skill via Codex/Claude code, training with analysis frameworks and methodologies from past experience. But when I use the skill with latest model to solve my casestudy, the output of Gen AI still cannot meet my expectation.
I used to be a strategy consultant focused on Auto, TMT, and Industrial Goods. From my personal view, the cabability of GPT and Claude in consulting work is still in the junior level, due to the limited industry data source and know-how accumulated from real business
GenAI product and AI agents are only able to finish basic work or called junior's work such as secondary desk research and expert interview notes taking (from transcript).