Use project management skills to write reminder words, the effect is explosive!

Although big models are getting smarter, there are still huge differences between people in how they use them. Some people use AI like running a company, and they can coordinate the operations of various departments with just a few words, which is efficient and fruitful. Some people use AI like teaching their grandma how to buy things online, and they use several different methods to teach her, but she still doesn't quite understand.

In the use of big models, the gap between people is mainly caused by the difference in the level of writing prompt words. Although there are now inference models, they can guess and infer what answers users want to a certain extent. However, inference models cannot replace prompt words. The current situation is that if you can write prompt words, AI is a powerful army, but if you can't write prompt words, AI is just your grandmother.

Interestingly, many senior project managers use the same ideas when directing teams, breaking down requirements, and managing stakeholders, which are also very useful in writing AI prompts. This gives project managers the opportunity to build their own moat in the AI ​​era .

Today we will talk about how to "transplant" project management skills into writing prompts, so that project managers can instantly become Prompt masters.

From goals to plans

When doing a project, project managers usually first clarify "what the project wants to achieve", and then break down resources, timelines, and key milestones. The same is true for writing prompts. Want AI to write copy? Give AI a concise and concise background of the requirements. Want it to do data analysis? Tell it the data indicators or inference ranges you care about.

for example :

“Please help me write a technical blog for junior programmers. The theme is the basic introduction of Python crawlers, focusing on practical cases and being simple and easy to understand. The length should be about 1,000 words.”

This way, the AI ​​can move towards a specific goal, rather than swimming in the vague pool of "technical blog". Many people start by asking the AI ​​to "write a technical blog" without giving any context, which results in the AI ​​only being able to respond with "the key points of the technical blog as I understand it", which is naturally less effective.

WBS Decomposition Requirements Thinking

The WBS (Work Breakdown Structure) often mentioned by project managers is actually very consistent with the "multi-round questioning" of writing Prompt. If you throw all the requirements to AI at once, you will get a hodgepodge; if you let AI output in segments, you can control the direction of the content more accurately.

Example :

1. 1. “Please help me list the common core elements of a market analysis report, such as background, target market, competitive landscape, etc.”
2. 2. "Next, for the above core elements, add the data that needs to be collected, the data sources, and the corresponding analysis methods."

Follow The process of decomposing tasks using WBS can allow AI to produce more complete results step by step, rather than just pushing it to it and letting it "figure it out on its own."

Clarify the stakeholders’ perspective and let AI “take on the role”

Project managers often need to identify stakeholders and understand their needs. This technique also applies when writing prompts. In order to make AI output more "human-friendly" answers, you can also guide it to "play" a certain role to think about the problem.

example :

“Imagine that you are a senior short video operations manager. Write three operation strategies for novice creators on Douyin and provide corresponding cases.”

The reason for role-playing is that AI (especially models with reasoning capabilities) can give more professional and scenario-based answers based on the knowledge, tone, and focus of a specific identity; it will actively mobilize existing "operational experience" for reasoning, rather than just blindly copying the so-called "standard answers."

Establishing acceptance criteria

In the project, we will write acceptance criteria (Definition of Done) for each requirement to ensure that everyone knows clearly when it is completed. The same is true for writing prompts. You have to let the AI ​​know: what kind of output meets the requirements.

refer to :

Please help write a new feature requirement document, which must include:

1. 1. Background and objectives;
2. 2. User pain points and demand analysis;
3. 3. Feature list and prioritization;
4. 4. Risks and countermeasures. Please use the Markdown heading structure, and each section should be at least 100 words.
   

Like this, you give AI a checklist and let it fill in the "requirements list" one by one. Ordinary AI may only give you a general idea, but after parsing these specific standards, the reasoning AI will be more conscious of the "points" output, and will not just pass by like before.

Iteration and feedback

When doing projects, especially in agile teams, we often use an iterative approach: first make an MVP (minimum viable version), and then continuously optimize based on feedback. Letting AI write things can also be done in multiple rounds of iterations:

• • First let AI come up with a first draft;
• • Then provide revision suggestions;
• • You end up with a piece of content that better matches your expectations.

for example :

• • First question: "Write a microblog copy calling for environmental protection, with a lighthearted and lively tone."
• • After receiving the first draft: "Shorten it a bit, keep it under 50 words, and add a closing slogan with 'take action now'."
• • If it’s still not enough, polish it further: “Please use words with a sense of urgency to make people feel the severity of environmental pollution.”

Reasoning AI will often remember your feedback and requests in multiple rounds of conversations and automatically make logical corrections. If it is a normal AI model, it may lose the context after one conversation and have to repeat your request over and over again, which is not only inefficient but may also give irrelevant answers.

Priority Management

Some requirements are "must haves", some are "optional", and some are "unnecessary". In order to prioritize requirements, project managers will use the MoSCoW (Must have, Should have, Could have, Won't have) principle. When writing prompts, you can also list the "key" and "optional" information first, so that AI knows which parts should be focused on.

Example :

“Help me write a new product requirements list:

• • Must have: core search function, user login function;
• • Should have: Collection and sharing functions;
• • Could have: Personalized recommendations;
• • Won’t have: Payment functionality.

In this way, AI can first explain the "Must have" part thoroughly, and then take into account other needs. If no priority is given, AI often writes too scattered in order to appear "comprehensive", and the real key points are drowned in a lot of information.

Risk Management

Identifying risks is a skill that project managers are very skilled in. When communicating with AI, you can first point out in the prompt words which aspects you are particularly concerned about or prone to problems, so that AI can be prepared and make more careful inferences.

for example :

“I want a smart home market analysis report, but please note:

1. 1. Do not quote unverified market data;
2. 2. If you have a prediction, please base it on logical deduction;
3. 3. Do not copy any confidential content from industry reports.

In this way, when AI generates content, it will pay more attention to the legitimacy of the data and the reasoning process, making the answers more reliable. Otherwise, some models may directly "fabricate" some data, or copy a bunch of reports without any source, which can easily make you fall into the "pie in the sky trap".

Last words

We all know that AI is a double-edged sword. It can greatly improve work efficiency, but it may also give "parallel answers" when there are no constraints. When you start to manage Prompt with the mindset of a project manager and work with reasoning AI, you will experience the feeling of "effective communication and smooth execution".

You never expected that the first people to land on the shore in the AI ​​era are project managers. Congratulations to all project managers!