Last week, a trending post on Juejin pointed out a detail: someone installed twenty or thirty AI plugins in a week, and when it came time to actually do work, couldn't recall a single one to use. The post's verdict: knowing what you've installed is the first gate of onboarding; knowing where your work is stuck is what actually counts as crossing the threshold. Play with AI long enough, and what you're training isn't tool knowledge — it's diagnostic ability.
What this is
The post offers five test questions. Answer yes or no by gut reaction, no looking things up: in the past month, was there a piece of work you handed entirely to AI and got the result back; is there a "do it three times a week" task you've already handed to a fixed workflow; when you delegate, do you add a line like "who's the reader, what format, what to avoid"; in the answers it gives you, have you caught any errors; when it gets stuck, can you say whether it's missing context, perspective, or tools. Zero or one "yes" means you're still doing it the old way; two or three means you can use it but every task needs fresh briefing; four or five means you no longer care about tool count, only about where the work is stuck.
The post groups AI's weak spots into three categories: missing context (it doesn't know what your business looks like), missing perspective (answers read like correct-sounding filler), missing tools (it can't reach the data). The terms Skills (plugins that help AI do a task by template), role experts, and MCP (the protocol that lets AI read your company's internal data) — recognizing them is enough, no need to memorize. What stays valuable is the question itself: what exactly is it missing.
Industry view
This take resonates inside the tool-education community. Objections also exist: some feel the post underestimates the disruptive power of new tools themselves. Two years ago, no one imagined AI reading your ticketing-system database directly; tomorrow's new tools may turn today's "diagnostic ability" into redundant basics. There's also a more grounded critique: diagnostic ability sounds right but is nearly impossible to teach ordinary employees who've never touched business modeling — it's like telling a new hire "have judgment" without saying how to train it. Our view: the direction is right, but one piece is missing — diagnosis requires first having a list of "which of my work is repetitive labor," which is precisely the part most people are too lazy to write down.
Impact on regular people
For enterprise IT: where money should really go shifts from "company-wide tool training" to "helping business staff articulate their sticking points clearly" — IT department training content needs a redesign.
For individual careers: list everything you did this week, find the most repetitive one, see whether it's missing context, perspective, or data access — then decide whether to bring AI in.
For consumer market: AI tools' selling point shifts from stacking features to "helping you think through the problem" — products that design a strong onboarding questionnaire will have more stickiness than those with prettier feature lists.