Beyond the Hype: The Limits of AI in Project Management

Beyond the Hype: The Limits of AI in Project Management

Dr. Amir Fard Bahreini

Start

Self-Reflection

Q1

Q2

Q3

How have you used AI tools (ChatGPT, Copilot, Midjourney, etc.) in your personal life or studies?

What went well? Did it save you time, give you new ideas, or make your work easier?

What went poorly? Did you ever get wrong, misleading, or unusable results?

Q4

Did you ever feel tempted to rely too much on AI instead of doing the work yourself?

Class Notes

What Went Well

• Helped shortlist and narrow down ideas
• Made day-to-day work more efficient
• Generated itineraries, outlines, and shopping lists
• Explained Excel formulas and supported coding tasks
• Removed barriers to entry and boosted organization
• Summarized short documents and handled routine “time-waster” tasks

What Went Poorly

• Too agreeable / “Yes Man” tendencies
• Sometimes unclear or forgetful
• Struggled with understanding certain questions
• Risk of information overload or fake sources
• Formatting presentations was weak
• No access to company-specific or private info
• Occasional plagiarism concerns
• Overreliance Patterns
• Heaviest use during busy periods (exams, polishing work)
• Often leaned on for routine, low-stakes task
• Sometimes split across platforms (ChatGPT, Co-Pilot, Mayo)
• Rarely used for the most important decisions

AI in Project Management Today

AI offers compelling toolsbut the reality often includes hidden risks

AI can play various roles in PM

AI automates scheduling, resource allocation, and routine status reports

Helps with forecasting costs, identifying risks, and drafting progress updates.

Reality (What often happens)

Promise (What AI sells us)

• Predictions fail when the data is incomplete or messy.
• Recommendations lack human context and can damage stakeholder trust.
• AI sometimes delivers confident but wrong answers.

• Faster insights than manual spreadsheets.

• Reduced workload for project managers.

• “Data-driven” decisions that appear objective and rational.

Limitation 1

Data Quality Issues

• Garbage in, garbage out (GIGO): If project data is flawed, AI cannot magically “fix it.” It will amplify the error.

• Common issues: Missing timesheets, incomplete cost data, inconsistent naming conventions.

• Hidden danger: AI outputs still look professional, even if they are wrong, making it easy for managers to trust them blindly.

Additional Reading: DATA GARBAGE IN – AI GARBAGE OUT

Limitation 2

Lack of Context

• AI cannot read organizational culture: It does not see the political pressures, stakeholder relationships, or reputational risks that shape real decisions.

• Team dynamics ignored: AI may recommend assigning more work to a team member with “capacity,” but it cannot detect burnout, morale issues, or hidden conflicts.

Additional Reading: Missing Context: Understanding the Limitations of Artificial Intelligence

Limitation 3

Overconfidence & Hallucination

• AI sounds confident, even when it’s wrong. You may often trust it because of its professional tone.

• Hallucination risk: AI may invent KPIs, suggest resources that don’t exist, or cite non-existent case studies.

Additional Reading: Hallucination (artificial intelligence) Wiki

Limitation 4

Algorithmic Bias

• Bias in data = bias in outcomes. If past projects reflect biased resource allocation, AI will replicate those patterns.
• Hiring and staffing risks: AI might unfairly allocate tasks based on biased historical patterns (e.g., assigning fewer leadership roles to certain groups).

Additional Reading: What is algorithmic bias? by IBM

Limitation 5

Security & Ethical Risks

• Data leakage risk: Feeding sensitive budgets or project schedules into public AI systems (like ChatGPT free version) can compromise confidentiality.
• Accountability gap: If an AI system makes a flawed decision, who is responsible—the AI vendor or the PM who used it?

Additional Reading: AI Challenges by UT-Dallas

Limitation 6

Dependency & Skill Decay

• Overreliance erodes skills: If PMs always outsource tasks like critical path analysis or risk prioritization, they lose the ability to do them manually.

• Students may graduate unable to do “basic PM math” without AI help.

Alavi, M., & Leidner, D. E. (2001). Knowledge management and knowledge management systems: Conceptual foundations and research issues. MIS quarterly, 107-136.

Limitation 6

Dependency & Skill Decay

https://www.forbes.com/sites/dimitarmixmihov/2025/02/11/ai-is-making-you-dumber-microsoft-researchers-say/

https://time.com/7295195/ai-chatgpt-google-learning-school/

How to Properly Use AI

• Use AI as a Co-Pilot, Not the Pilot
• Always Validate the Data
• Keep Control of Context

• Guard Against Dependency
• Protect Your Data & Ethics
• Stay a Producer, Not Just a Consumer

Get Better at Prompt Engineering

The Importance of Prompt Engineering

• AI only gives answers as good as the questions you ask.

• A vague prompt = vague, generic results.

• A clear, structured prompt = specific, useful, actionable output.

A good prompt will help you

• Control quality → reduces risk of “hallucinations” and irrelevant answers.
• Save time → fewer follow-ups and re-dos.
• Stay critical → forces you to clarify what you really need before asking AI.

Additional Reading on prompt engineering is available on Canvas.

Final Note

AI isn’t free. it’s for-profit, and if you’re not careful, it can trap you in the role of a forever-consumer.

Final Note

• In this class, we have guidelines on how AI may be used. But ultimately, the decision of how you use AI in your life is entirely yours.
• The only truly bad decision is the one you make without understanding its consequences, both the obvious risks and the hidden ones.
• Use AI thoughtfully, not blindly. Let it be a tool, not a sole decision maker.

Thank you