Trust the Data - but Not Blindly: An Ethics Bistro on AI

Email Notifications

SHARE
Request to reuse this

It was a rainy Tuesday when the red flag popped up. The AI tool, designed to optimize resource allocation across our project portfolio, had flagged three critical projects for delay. The model’s recommendation? Shift half the team from Project Titan to Project Eclipse to balance out workloads.

At first glance, it seemed logical. The resource allocation maps, and velocity graphs supported the reallocation. But something did not sit right.

I had collaborated closely with Titan’s team leads for months. They were on the verge of a breakthrough with a critical client deliverable. Moving people now, even with Eclipse falling behind, could cause a domino effect across our most valuable account.

I called a huddle.

“Why did the model deprioritize Titan?” I asked the AI SME.

“It is based on risk scoring from delivery variance, budget utilization, and resource burn. Titan looked stable, so it pulled from there.”

“But it does not know the client conversation we had last week. Titan’s ‘stability’ is built on momentum we cannot afford to interrupt.”

That was it. The AI had the data but not the context.

We chose not to follow the recommendation. Instead, we manually adjusted scope and brought in temporary support for Eclipse. It was a tough call, but three months later, Titan delivered on time and exceeded client expectations. Eclipse caught up too—without derailing the portfolio.

That experience taught me something: AI is brilliant at pattern recognition, but it does not see what you know. It does not read nuance. And it does not carry responsibility.

So, when should project managers trust AI—and when should we intervene?

Trust AI when:

• You need unbiased, data-driven insights fast.
• The decision space is clearly defined and repeatable.
• You are analyzing trends across massive datasets where human bias or oversight might creep in.

But intervene when:

• The stakes involve human relationships, trust, or reputational risk.
• The model’s logic lacks access to critical context.
• The recommendation “feels wrong” and your intuition is backed by experience not fear.

AI is like a junior analyst with infinite memory and no emotional baggage. But it lacks judgment, and judgment is where leadership lives.

As project managers, we are not just responsible for outcomes; we are stewards of values. According to the PMI Code of Ethics, we are bound to act with responsibility, respect, fairness, and honesty. Blindly following AI no matter how accurate without human oversight may compromise all four.

Use AI like a compass not a map. Let it guide your thinking, but do not let it override your wisdom.

Because when things go south, the algorithm will not be in the room explaining the outcome you will.

So next time your AI flags a decision, pause. Ask: Does this align with what I know, what I have seen, and what matters most?

If the answer is no, trust yourself and intervene.

Reference:

PMI Code of Ethics

Webinar: Ethical Project Leadership in the digital age

Webinar: When to Trust AI and When to Intervene