Most AI training teaches the tool. Almost none teaches the work.
You can find a hundred courses on what a large language model is. Far fewer will sit with you while you do the thing you actually have to finish on Thursday.
What is wrong with how this is taught now.
Two problems keep showing up. The first is that courses stop at the demo: you watch someone summarise a document, and then you are on your own with your document, which is messier. The second is that almost nobody teaches you to check the output. Courses will happily walk you through producing a budget, a grant application or a board update, and then never mention how you would know whether the numbers are right.
4 LLMs (e.g., Claude, ChatGPT), and what each is good at.
| Capability | Claude | ChatGPT | Gemini | Perplexity |
|---|---|---|---|---|
| Long documents, 40 pages and up | Yes | Partly | Yes | Partly |
| Sources you can click and check | Partly | Partly | Partly | Yes |
| Careful, restrained long-form writing | Yes | Partly | Partly | No |
| Works inside Docs, Sheets and Gmail | No | No | Yes | No |
| Reads files you upload | Yes | Yes | Yes | Partly |
| Connects to your other apps | Partly | Yes | Partly | No |
| Free tier good enough to learn on | Partly | Partly | Partly | Yes |
YesPartlyNo
No vendor will tell you a competitor is the better choice for the job in front of you. That is the gap this fills.