WiserWulff turns 30 just as the business world reaches an inflection point. This is a opportune time to reflect back on the last generation(s) of project management and use it to inform where it is going in a post-AI world.
This series covers what’s changed, what hasn’t yet changed, and what will remain a constant.
What’s Changed with Requirements Management
Project management has gone through two distinct generations over the last three decades.
First, waterfall gave way to Agile. Scrum, Kanban, SAFe, and other frameworks changed how teams approached delivery. Technology transformed how projects are planned, executed, and monitored.
Now, AI agents are changing how work gets done.
Yet fundamental principle remains unchanged: Successful projects begin with clear requirements.
The tools and methodologies may change. The need to define what a project must accomplish does not.
What Hasn’t Changed with Requirements Management
Every project exists for a reason. There is a problem to solve, an opportunity to pursue, or a business need to address. Before a team can determine how to deliver a solution, it needs to understand what that solution needs to accomplish.
That is where requirements come in. Requirements provide the foundation for the work that follows. They help define the desired outcome, establish the conditions the solution needs to meet, and give the team something against which to evaluate the finished product.
The methodology does not change that. Agile did not make requirements unnecessary. Different methodologies provide different approaches for managing the work, but every successful project still needs a clear understanding of what it is trying to achieve. The requirements do not need to account for every possible detail before work begins: Projects change. Priorities shift. Teams learn things during execution that they could not have known at the start.
The point is not to eliminate flexibility; now more than ever, the point is to establish a strong enough foundation that those changes don’t make you lose the objective.
…Insert AI
The rise of AI agents in project management introduces a significant change to project delivery. For decades, much of the work of turning requirements into deliverables was performed by people. Increasingly, AI can now perform many of these activities: Agents can analyze information, generate content, write and review code, produce deliverables, and carry out ever-more complex workflows. As these capabilities improve, more of the execution layer will be handled by AI.
This makes requirements even more important. A human team can often compensate for an incomplete requirement. Someone can ask a question, recognize an inconsistency, or use experience to understand what was probably intended.
While AI agents can also reason and make decisions, they are still operating within the objectives, information, instructions, and constraints they are given. If those are incomplete, the agent has to fill in the gaps. That can create a problem when the missing information involves something fundamental to the project.
AI can make execution faster. It cannot make an undefined objective well-defined.
Prompts Are the New Requirements
A prompt tells an AI system what you want it to accomplish. Depending on the task, it can also provide context, constraints, inputs, priorities, and expectations for the final output, much like what you would see in a project charter or business requirements document (BRD).
When many think of prompting AI, they imagine providing a one sentence question, or perhaps, one paragraph request to a chatbot. But a complete set of instructions for an agent should much more like traditional project requirements, including an entire set of business rules, technical specifications, and acceptance criteria. This establishes the 5 W’s of a project before the work is performed.
In traditional development, requirements define the expected outcome so a team can build the right solution.
In an agentic environment, prompts and other instructions define the expected outcome so an AI agent can perform the work.
The technology is different. The underlying discipline is not. The better requirements, the better the foundation for execution.
The More AI Does, the More Management Is Necessary
This shift has implications for the skills organizations need. If AI handles more of the execution-oriented work, people can spend less time performing time-consuming, repetitive and error-prone tasks. More of their time can go toward defining the problem, establishing the desired outcome, managing requirements, and evaluating the results.
That makes requirements management a particularly important skill for the future of work. Someone still needs to determine:
- What problem are we solving?
- What does the solution need to accomplish?
- What constraints need to be considered?
- What information does the agent need?
- What does a successful result look like?
- How will we know when the work is complete?
- Those questions existed long before AI.
AI simply gives them a new context.
The person defining the requirements may no longer be writing a specification for another person to implement. They may be defining the requirements that an AI agent will use to perform the work.
This creates a whole new discipline and set of core competencies that organizations need in order to rigorously define outcomes, constraints, assumptions, inputs, and acceptance criteria. The future of work will create all new roles, including:
- AI Delivery Leaders, translating business workflows into holistically implemented, integrated, and adopted agentic workflows that handle them end-to-end
- Process Masters (with indeed an entire team underneath them) to capture the entire organizational ontology and keep it up to date
- AI Quality Assurance Specialists, who serve as the agents’ “supervisor” to identify and manage errors, low quality outputs, and other issues inherent to agentic work
- Agent Developers, Engineers and Architects to address those issues, as well as capitalize on new opportunities
- An enhanced information security, privacy, and legal team managing the liabilities unique to an AI-augmented workforce.
These roles become especially important as AI agents become capable of performing increasingly complex work.
A Strong Foundation Still Leaves Room to Adapt
There is a temptation to assume that better requirements mean more detailed requirements. That is not necessarily the case. Good requirements provide enough clarity to establish the desired outcome and the conditions the solution needs to satisfy.
They do not need to predict every decision that will be made during execution. Projects will still uncover new information: teams will still identify better approaches; stakeholders will still change their priorities; AI may even identify possibilities that the project team had not considered. A strong foundation makes those changes easier to manage. However, without a clear understanding of the intended outcome, every change becomes another opportunity for confusion. With requirements in place, teams can determine whether a new idea improves the solution or simply creates unnecessary scope.
The same principle applies when AI is performing the execution. Give an agent a clear destination, and there is room to determine the best way to get there. Give it an unclear destination, then even perfect execution will not necessarily solve the problem.
Next Steps for Your Organization
Organizations looking to incorporate AI into project delivery should start by examining their requirements practices. Before investing heavily in new AI tools or agentic workflows, consider whether the organization can clearly define the work those tools are supposed to perform. Ask:
- Are project outcomes clearly defined?
- Are requirements documented and accessible?
- Do stakeholders agree on what success looks like?
- Are acceptance criteria clear and objective?
- Are important constraints and dependencies identified?
- Is someone accountable for requirements quality?
- Can AI systems access the information they need to perform the work?
- If the answer to several of these questions is no, AI may not solve the underlying problem.
It may simply allow the organization to accomplish the wrong objectives faster. That is not progress.
The organizations that get the most value from AI will not simply be the ones with the most advanced tools. They will be the ones that understand what they want those tools to accomplish and can clearly communicate it.
Contact the firm that has helped organizations plan and deliver successful projects since 1996.
