通用助手 vs 教师专用工具
什么时候一个空白对话框胜过一个任务化的按钮，什么时候相反。
本文尚未翻译成中文，以下正文为英文。
Two things are sold to teachers under the same "AI" label and they are not the same product.
A general assistant is a blank box. It will do anything you can describe, and the quality of what comes back is mostly a function of how well you described it.
A purpose-built teacher tool is a set of named jobs — write a rubric, level this text, draft a parent email — each with a form instead of a prompt.
Where the purpose-built tool wins
Repeated, well-defined jobs. If you write exit tickets every week, a button that produces exit tickets beats retyping the same prompt with slightly different words each time and getting slightly different structure back.
Colleagues who will not write prompts. This is the honest reason most schools buy one. The general assistant has a higher ceiling and a much higher floor of effort. A department where three people will invest in prompting and twelve will not is better served by forms.
Consistency across a team. Same tool, same shape of output, and a rubric that looks like every other rubric in the department.
Where the general assistant wins
Anything unusual. The moment your task is not on the menu, a task-shaped tool has nothing to offer and the blank box has everything.
Working from your own material. General assistants take long documents. Many task-shaped tools take a topic and a year group, which puts you straight back into generic-content territory.
Arguing. The most valuable thing in the whole workflow is being able to say "this is front-loaded, rebalance it" — and a form has nowhere to put that.
The decision, compressed
| If your situation is… | Reach for | |---|---| | The same job every week | Purpose-built tool | | A whole department, mixed enthusiasm | Purpose-built tool | | Working from a specific text or document | General assistant | | A one-off, awkward task | General assistant | | You want to push back on the first draft | General assistant |
The thing neither solves
Neither category knows your class. Both will produce something plausible for a class that does not exist, and the plausibility is the danger — a bad lesson plan that reads badly gets rewritten, while a bad lesson plan that reads well gets taught.