TutorialsEveryone a Product Manager: Build Your Own Tools with AI
Six questions, one checklist, one working first version — you make the judgement calls, AI writes the code, and you verify it yourself at the end.
This course is not about writing code. It is about making the request clear before you hand it to AI. Pick one small annoyance from your own week, answer six questions in a Word checklist, hand it to Baidu Dumate for a first version, then verify it with an example you wrote yourself. No code is involved at any point.
What you take away
| Deliverable | What it is |
|---|---|
| A way to choose the problem | Pick one concrete, small, checkable annoyance from the past week instead of a vague big idea |
| One checklist | A Word form with six questions. Tick the common options, fill in a few lines, and you have a brief AI can read |
| One collaboration prompt | Paste it to Baidu Dumate: it owns interface, interaction, technical approach and code; you own judgement and acceptance |
| One working first version | A tool you can open and operate, plus an acceptance table — not a mockup, something you actually use once |
If you are stuck on these
- Too many ideas: the moment you describe a tool you also want login, cloud sync and Excel export, and nothing gets built;
- Can not describe it: all you can tell AI is "make me a list-cleaning tool", and what comes back is not what you had in mind;
- Afraid to trust the result: the page opens, but you have no way to tell whether the output is right;
- Built once, then abandoned: never used for real work, so you never learn what to change next.
What you will be able to do
Pick a problem you can finish
Four questions filter out ideas that sound impressive but cannot be built in one sitting.
Describe a tool AI cannot misread
Three lists and six questions pin down who has what problem, what version one does, and what counts as correct.
Verify it for real
Write the expected result first, then test. When it is wrong, report input, expectation and actual behaviour so AI can fix exactly that.
Who this is for
Course structure
Lesson 1 · Find the problem: what natural-language programming really is, how to source a topic from your own week, and how to size version one.
Lesson 2 · Fill in the three lists: two questions each for requirement, design and acceptance, filled in a Word checklist, plus a worked acceptance example.
Lesson 3 · Build and verify: install Baidu Dumate, hand over the checklist and prompt, decide the open questions one at a time, then test the first version yourself.
Course materials
Four files, all downloadable:
- AI实用工具-三张清单模板.docx — the Word checklist used in class, with the full collaboration prompt on its last page
- AI实用工具-制作提示词.docx — the prompt that starts the build, plus two ways to phrase a change request
- AI实用工具-课程讲义.md — the full handout, including six self-study prompt templates and blank document templates
- AI实用工具-课件V4.pptx — the 38-slide deck used in class
The checklist is a Word form: open it in desktop Microsoft Word with editing enabled to tick the boxes.
The one principle
Everything in this course comes down to: AI implements, you judge. Product goals, scope trade-offs, which open question to pick, whether the result is actually usable — these cannot be delegated. Layout, interaction flow, technical approach, code and tests — these can.
Three red lines before you start:
- Check the rules yourself — test filtering, deduplication and sorting against an answer you already know, not against AI saying "done";
- Anonymise your material — use fictional or cleaned samples; never paste credentials, keys, customer data or internal documents, and keep them out of screenshots;
- Write "untested" when untested — a page that opens is only step one; anything you did not actually run stays marked as untested.
From Benchmark Dissection to Original Copy
Parse one benchmark creator's five scripts with the Social Media Assistant, break down and iterate the style prompt, then write an original piece of your own.
Finding the Problem: Start From One Concrete Annoyance
What natural-language programming actually is, how to source a topic from your own week, and how to decide how much version one should do.