The greatest value of GenAI is not that it can produce polished, copy-and-paste-ready content for us. Its real value lies in helping us discover the questions we did not ask, the assumptions we did not see, and the true business purpose behind the stated requirements.

For many project managers, the first instinct when using AI is to feed it a long document and ask it to summarize, rewrite, generate a PRD, create user stories, or produce presentation content. In software projects, some also start to see AI as an accelerator for writing code, producing documents, and creating slides. Of course, these use cases are useful.

But if AI is only helping us repackage existing content, then its real value has not yet been fully realized.

I also use AI quite often for requirements analysis, key-point summaries, concept mapping, creative thinking, and even for producing documents or presentation content. But I usually do not expect AI-generated content to become the final answer directly. Accuracy is still my responsibility.

The value of AI is not to replace judgment, but to expand my field of observation. It can help me look at an issue from different angles, identify contradictions, surface assumptions, and generate follow-up questions. As for which ideas are valid and which are not, I still need to go back to the context, the facts, and the stakeholders for further confirmation.

This is how AI helps me avoid being limited by my own experience and knowledge, and reduces the blind spots created by that experience.

This idea is especially relevant when applied to project requirements analysis. Requirements analysis begins at the start of a project, and often even before the project officially starts. If the requirements are not clarified properly, then no matter how beautiful the downstream planning looks, the project will often end up with a painful case of “garbage in, garbage out.”

Following the general PMI/PMI-PBA context, requirements work typically involves steps such as elicitation, analysis and modeling, definition and refinement, acceptance criteria definition, validation and verification, and requirements management and traceability.

But in practice, the real challenge is often not knowing how to perform these steps. The real challenge is discovering the problems that have not yet been spoken out loud at each step.

This ability to uncover the requirements gaps that we ourselves did not see is where AI can create the greatest value.

For example, many requirements specifications may include a requirement like this: “The system should allow users to quickly check the status of a case.” At first glance, this seems to be a simple case-status query function. It does not look very complicated. Many people may only ask questions such as, “Which fields should be displayed on the page?” or “Should the searchable cases be filtered based on user permissions?”

But if we look deeper, there are many additional questions we could ask:

  • Who needs to perform the search? Managers, case handlers, customer service staff, or external citizens?
  • Does “quickly” mean within three seconds, or does it mean reducing the number of operation steps?
  • Should the data reflect real-time status, or is daily batch updating sufficient?
  • Should the system keep query logs? If yes, for how long?
  • How should the screen respond when the data is incomplete?
  • Is this requirement really trying to solve an efficiency issue, a transparency issue, or an accountability issue?

Through these questions, we may discover that behind the seemingly simple “query function” lie issues related to permissions, data updates, accountability boundaries, user roles, and acceptance criteria.

If we provide AI with more context, such as background information about the project, the customer, the business process, or the constraints, AI can often help us identify even more requirements gaps.

In general, the requirements gaps that AI can help us uncover can be grouped into three categories:

- Problem Gaps:

Are we asking the wrong question, or only seeing the surface-level problem?

- Context Gaps:

Are we missing roles, workflows, constraints, or exception scenarios?

- Acceptance Gaps:

Are the requirements verifiable, testable, and traceable?

From these three perspectives, AI can help us generate follow-up questions, identify root causes, supplement roles, scenarios, and constraints, and further transform requirements into acceptance criteria or testable scenarios.

In this way, AI is no longer just a token-hungry document assistant. It becomes an intelligent partner that can truly help project managers improve the quality of their judgment, expand their field of observation, and create greater value in their work.