A practical guide to where AI helps project managers—and where it falls flat.
After using AI tools daily for the past few years, I’ve developed a clear picture of where they add value and where they waste time. This isn’t about what AI could do theoretically—it’s about what actually works in day-to-day project management.
5 Tasks AI Does Well
1. Meeting Notes and Action Items
Why it works: Meeting notes follow a predictable structure—attendees, topics discussed, decisions made, action items. AI excels at turning messy transcripts or rough notes into clean, organized documentation.
How I use it: I feed a meeting transcript (or my rough notes) into AI and ask for:
- Summary of key discussion points
- Decisions made
- Action items with owners
- Open questions for follow-up
- Topics and notes in table format
Suggested prompt: “Create meeting notes from this transcript. Include: action items with owners, a table of topics and details, and decisions made (what was decided and who made the decision).”
Time saved: What used to take 20-30 minutes of post-meeting cleanup now takes 5 minutes of review and light editing.
2. First Drafts of Routine Documents
Why it works: Project charters, status summaries, stakeholder updates—these documents follow patterns. AI can generate a solid starting point that you refine with your specific context.
How I use it: I provide the key information (project name, objectives, stakeholders, timeline) and ask for a structured first draft. I always review and revise, but I’m editing instead of writing from scratch.
Enterprise AI advantage: With M365 Copilot, I search my email and Teams discussions to gather details for weekly status reports. Instead of manually hunting through threads, Copilot finds the relevant updates.
Suggested prompt for status reports: “Search my emails and Teams messages from the past week related to [Project Name]. Find: completed work, blockers or issues raised, upcoming milestones mentioned, and any decisions made. Summarize these into a weekly status report format with sections for Accomplishments, In Progress, Blockers, and Next Steps.”
Important caveat: Keep documents small. AI handles one-page summaries better than 20-page project plans. For larger documents, work in sections.
Time saved: A first draft that would take 45 minutes to write takes 10 minutes to generate and 15 minutes to refine.
3. Brainstorming Risks and Issues
Why it works: AI is good at generating comprehensive lists. For risk identification, it can suggest categories and specific risks you might not have considered.
How I use it: “What risks should I consider for a [type of project] in [industry/context]?” Then I filter the list based on what’s actually relevant to my organization.
Suggested prompt: “I’m managing a [software migration / system implementation / process change] project for [industry context]. Generate a risk register with categories including technical, organizational, resource, schedule, and external risks. For each risk, include: risk description, potential impact, likelihood (high/medium/low), and suggested mitigation strategies.”
What AI provides: Breadth. It generates a wide range of possibilities.
What I provide: Depth. I know which risks actually matter for this project, this team, this organization.
Time saved: Risk brainstorming sessions start with a comprehensive list instead of a blank whiteboard.
4. Translating Technical Content for Different Audiences
Why it works: AI is effective at adjusting complexity and tone for different readers. Technical specifications become executive summaries. Jargon becomes plain language.
How I use it: I paste in a technical document and ask for a summary appropriate for [specific audience]. For example: “Summarize this for a non-technical executive who cares about timeline and budget impact.”
Time saved: Translation work that used to require multiple drafts happens in one pass.
5. Preparing for Difficult Conversations
Why it works: AI can role-play the other side of a conversation, helping you anticipate objections and prepare responses.
How I use it: “I need to tell the project sponsor that we’re two weeks behind schedule. What questions will they likely ask, and how should I address them?” Or: “What objections might the CFO raise about this budget increase?”
Time saved: Preparation time stays the same, but preparation quality improves significantly.
5 Tasks AI Does Poorly
1. Anything Requiring Organizational Context
Why it fails: AI doesn’t know your organization’s history, politics, or unwritten rules. It doesn’t know that the last three projects with Vendor X went poorly, or that the VP of Engineering and VP of Product don’t get along, or that “approved by the steering committee” actually means “approved by the CEO’s chief of staff.”
The risk: AI advice sounds reasonable but is completely wrong for your specific situation.
What to do instead: Use AI to generate options, but apply your organizational knowledge to evaluate them.
2. Stakeholder Relationship Management
Why it fails: Project success depends on trust, influence, and relationships. AI can’t read a room, sense tension, or know when to push back versus when to let something go.
The risk: Following AI-suggested “stakeholder engagement strategies” that ignore the human dynamics that actually determine project outcomes.
What to do instead: Use AI to prepare for stakeholder conversations, but rely on your own judgment in the moment.
3. Making Judgment Calls
Why it fails: AI can present options with pros and cons. It cannot tell you which option is right for your project, your constraints, your risk tolerance.
The risk: Outsourcing decisions to AI that require human judgment and accountability.
What to do instead: Use AI to expand your options and think through implications, but own the decision yourself.
4. Handling Confidential Information
Why it fails: Consumer AI tools are not secure environments for sensitive data. Anything you paste into a chat window could be used for training or stored in ways you don’t control.
The risk: Exposing client data, financial information, employee details, or strategic plans.
What to do instead: If your organization has an enterprise AI agreement with proper data handling, use that. Otherwise, anonymize information before using AI tools, or don’t use them for sensitive content at all.
5. Knowing When It’s Wrong
Why it fails: AI presents everything with equal confidence. It doesn’t flag when it’s guessing, when information might be outdated, or when it’s filling gaps with plausible-sounding nonsense.
The risk: Trusting AI output without verification, especially for factual claims or technical details.
What to do instead: Verify anything that matters. Treat AI output like work from an intern or assistant—helpful, but you still need to review it before it goes out.
One more thing: When AI coding assistants (e.g., Claude, Cursor) ask for permission to do something—run a command, edit a file, access something—make sure you understand what it wants to do before you approve. Read the request. If you don’t understand it, ask for clarification. AI assistants are powerful tools, but you’re still the one responsible for what happens.
The Pattern
Looking at these lists, a clear pattern emerges:
AI is good at: Structure, patterns, breadth, first drafts, translation, preparation.
AI is bad at: Context, relationships, judgment, confidentiality, self-awareness.
In other words: AI handles the parts of project management that follow predictable patterns. It struggles with the parts that require human understanding, judgment, and trust.
That’s not a limitation to work around—it’s a feature. The valuable parts of project management remain human. AI just makes the routine parts faster.
How to Apply This
Before using AI, ask yourself:
- Does this task follow a predictable pattern? (If yes, AI can probably help)
- Does it require organizational context? (If yes, verify AI output carefully)
- Does it involve confidential information? (If yes, consider not using AI)
- Does it require judgment about people or politics? (If yes, use AI for prep only)
The goal isn’t to use AI for everything. The goal is to use AI where it helps and skip it where it doesn’t.
Related Posts
- I’m a PM Who Uses AI Daily—Here’s What I’ve Learned
- Coming in Week 3: Your First Week Using AI as a Project Manager
AI is a tool, not a replacement. The project managers who understand where it helps—and where it doesn’t—will be more effective than those who either avoid it entirely or try to use it for everything.