Raw LLM Responses

People forget though that they're testing this on a generalized model used to fit general tasks. An easy solution to your problem for example is if someone just trained the AI specifically in the domain you're working in. Ie, training it more on data with inventroy management and decision making in that space, and fine tuning it - giving it thousands of datasets of different use cases, procedures, etc. A fine tuned proprietary model like that would be able to be created TODAY and not in the future because the technology exists. You'd just need to hire an AI engineer, pay the initial costs and it would be done and running a lot better than the general AI models used today. A good example of this is basically Aladdin - Black Rock's own prioprietary AI model which many don't know about but has been in existence since the 80's. Since the advent of more recent LLM technology, they've most definitely finetuned Aladdin and you can get they are using it for stock market predictions which probably have yielded them billions of extra dollars. Other examples are the medical field's use of their own trained AI which is more research data & study focused and used to pull peer reviewed information with high accuracy. When you're using a general public model, it will not be as accurate because it is trained to do everything rather than specificity and specialization. You can bet though, if it is fine tuned for specific tasks with proprietary data it will do significantly better. And lots of companies are already doing this because they realize the potential & the amount of $$ they can save in the future when they fire all their workers.