Five Ways to Strengthen Human Judgment in AI-Assisted Projects

AI is increasing the speed of project work. The value those projects deliver will depend on whether organizations can raise the quality of human judgment, learning and coordination at the same pace. Speed on its own delivers very little.

Why? Because AI’s outputs are based on the context available to it. In projects, some of the most consequential context is incomplete on unavailable in a digital form that AI can digest. This could include an unstated or still-emerging sponsor's priority, a regulator's shifting posture, or a stakeholder who may be withholding support due to unclear priorities. Project professionals surface that context through relationships, observation, and day-to-day interaction. Their judgment is what turns an AI-informed recommendation into a decision that can hold inside the organization.

Complexity compounds the issue. PMI's Pulse of the Profession® 2026 found that 81% of project professionals report projects have grown more complex in recent years. Complex environments are characterized by the very context that remains invisible to a model: interdependencies between initiatives, competing agendas across functions, and second-order effects that surface in conversation long before they appear in a status report.

Navigating this modern project environment was the theme of a recent PMI webinar I hosted, Built for Complexity: How AI Is Redefining Project Success, with Professor Johan Roos of Hult International Business School and Luxembourg School of Business, Dr. Leon Herszon of Value Edge Group and Ronald van Loon of Intelligent World.

Their perspectives converged around five practical guidelines for project professionals in applying human judgment to AI outputs.

1. Filter for what matters when information is abundant

Roos framed the shift that runs through the discussion. "For most of my career, the scarce resource was information," Roos said. "Today, information is instant and almost free. What has become scarce is practical wisdom — knowing which answer is worth acting on." His summary of the current moment is that we have never had more information and never been less sure what to do with it.

Organizations once struggled to obtain enough information to decide. AI now generates volume on demand, which moves the difficulty to filtering and establishing what deserves attention, what holds up under scrutiny and what connects to strategy.

Project professionals add value here by helping teams distinguish signal from noise, assess the credibility of AI-generated output, weigh trade-offs and convert analysis into coordinated action.

Practical emphasis: Treat AI output as an input to a decision.

Beyond asking What does the AI recommend? teams should ask:

• What assumptions underpin the recommendation?
• What context is missing?
• Whose perspective is absent?
• What follows if we act on this, and what follows if we do not?

2. Test the decision against the system it has to survive

"Project managers are no longer just managing projects," Herszon said. "They are managing continuous change." Projects sit inside interconnected systems shaped by technological, regulatory, economic and organizational shifts that arrive together and act on one another in ways that resist prediction. As he put it: "Complexity is not an exception. It's the default that we are seeing in the world today."

PMI's 2026 Pulse of the Profession shows how far that condition has spread. 97% of project professionals managed at least one complex project in the past year. Professionals that handle complexity well are roughly five times more likely to see their project succeed, and what distinguishes them is systems thinking (treating the project as a web of interdependencies rather than a task list), combined with strong business acumen. However, only 23% of project professionals currently identify systems thinking as critical to navigating complexity.

Faster analysis will not resolve unclear priorities, weak alignment, fragmented decision-making or burdensome governance. Project professionals must take up the role of sensemakers, connectors across stakeholder groups, and stewards of value.

Practical emphasis: Assess an AI recommendation at system level before acting on it.

• Which teams, stakeholders or systems will be affected?
• What other changes are underway at the same time?
• Could changing one part of the system create problems elsewhere?
• How easily could the decision be adjusted if conditions shift?

Herszon's teams use a shorter version at every meeting: What changed recently that can affect other parts of our project? The aim is sensing shifts early enough to adapt.

3. Surface the tacit knowledge behind your decisions

"For years, complex decisions in project management environments lived in people's heads," van Loon said. That knowledge was never documented, so it was never governed. As he put it, "it just worked, because humans are forgiving, flexible and contextual."

Experienced project managers know instinctively which stakeholder to approach first, which risk warrants immediate escalation, when a formal process should be adapted, which delay is survivable, which issue is about to become politically sensitive, and which trade-off senior leaders will accept.

AI cannot generate that kind of informal judgment. Van Loon watched this play out when a client tried to automate a "next best action" recommendation for project managers: "no two project managers were using the same criteria to make the same type of decision. The AI did not create this inconsistency. It revealed it."

Practical emphasis: When AI output looks incomplete or unrealistic, identify the experience a human is applying that the system lacks.

• What do experienced team members know that the data does not capture?
• Which informal relationships will shape how this decision unfolds?
• What criteria are people applying without stating them?
• Where would standardization help, and where would it remove necessary judgment?

4. Use AI to challenge decisions rather than make them

If AI produces answers quickly but cannot see the full organizational context, its strongest role is as a critical thinking partner. "Use AI to challenge your assumptions and your decisions," Herszon said. "Not to make them; to challenge them." He suggests putting the tool to work against the team's preferred position: attacking the favored option, surfacing overlooked risks, generating alternative interpretations of the same evidence, simulating stakeholder objections, and comparing courses of action.

Practical emphasis: Build challenge into the prompt and into the decision process.

• What assumptions would have to hold for this option to succeed?
• What evidence would disprove our preferred approach?
• Which stakeholders are most likely to object, and why?
• What second-order effects are we overlooking?
• Under what conditions would an alternative be stronger?

The project leader then weighs those possibilities against organizational reality.

5. Define governance before adding more AI

"Before you add more AI, define your governance model for it," van Loon said. Where AI influences project decisions, its role needs to be specified at the level of a particular decision or workflow rather than in a broad policy document.

Van Loon puts four questions to every organization he advises:

1. When can AI act?

2. When must a human intervene?

3. Who owns the decision?

4. How will the organization know whether AI improved the outcome?

These questions help distinguish between AI that informs, recommends, assists or acts. They also separate adoption from value: the measure is whether AI improves decision quality, speed, risk management or stakeholder outcomes, rather than how often the tool gets used.

Practical emphasis: Define decision rights at the workflow level.

• What role does the system play?
• What evidence must a human review?
• Who can approve or override the output?
• How will the decision be documented?
• What happens if the output is wrong?
• What result would show the AI added value?

This governance should be in place for any significant use of AI.

The bottom line: Match speed with judgment

The organizations that succeed with AI will be those that pair the technology with strong judgment, business acumen and the capacity to work in complex conditions. Technology can accelerate a decision. Ownership of it stays with people.

Project professionals are well placed to hold that line. They sit at the intersection of strategy, execution, stakeholders and outcomes, and they understand how a decision made inside a project travels across the wider organization. As AI is further embedded into daily work, the ability to connect context, judgment and accountability becomes more valuable rather than less.

The lesson I take from the conversation is that AI should strengthen critical thinking, or as Johan Roos puts it, our practical wisdom. When it produces an answer, the questions worth asking are what assumptions shaped it, what context is missing, how it affects the wider system, who remains accountable, and how the organization will learn from the outcome.

Project success will be determined by how effectively people combine technological capability with judgment, accountability and trust. Or, as our discussion concluded: start small, get the decision right, then go big.

Tags: Artificial Intelligence | Complexity | Leadership | Project Management | Problem Solving

Related Insights

You May Also Like