AI can help project managers prepare plans, organize meeting actions, examine risks, and draft status reports. A useful starting point is one specific task: define the information the AI can use, the output you need, and what must be verified before anyone acts.
One of the most important things I have learned working with AI is that a coherent answer can still be wrong. For me, a useful draft is only the starting point. It still has to meet the requirements.
In our software work, I have seen one AI model report that everything looked good, only for another AI model to find gaps. Our own review then found different issues again. That experience shapes how I approach AI in business: clear requirements, useful delegation, and checks that match the consequences of an error.
Project managers can apply the same approach.
Which project tasks are suitable for AI?
Start with a task where the source information is available and the result can be checked. PMI’s guidance considers both task complexity and the level of human intervention required when deciding how to use AI. PMI’s framework for AI in project management.
These five workflows provide practical starting points:
Workflow | Information to provide | Output to request | What to verify |
|---|---|---|---|
Planning | Scope, milestones, constraints, and known dependencies | Draft work breakdown and sequencing | Missing work, assumptions, capacity, and dependencies |
Risk review | Current risk register, issues, and project updates | Questions, possible risks, and draft responses | Supporting evidence, relevance, ownership, and priority |
Meeting follow-through | Authorized notes or transcript | Decisions, actions, owners, and unresolved questions | What was agreed, who accepted responsibility, and what remains unclear |
Status reporting | Dated project records and verified metrics | Draft update with exceptions and decisions needed | Numbers, reporting period, sources, and unsupported conclusions |
Project knowledge | Approved procedures and current documentation | Answers with source references | Document version, permissions, and escalation needs |
Which AI tools can support these workflows?
Start with the task and the output you need, then choose an approved tool that fits your workflow.
ChatGPT: draft, rewrite, and summarize content using supplied information. Check the output against your source records. ChatGPT capabilities overview.
Claude: work with uploaded PDFs and text documents to help review project material and prepare a draft. Confirm its interpretation against the documents. Upload files to Claude.
Microsoft Teams intelligent recap: prepare meeting notes and possible follow-up tasks from transcripts, subject to an eligible license, transcription, and organizational settings. Review decisions, owners, and dates. Microsoft Teams recap.
Check current licensing, data handling, and your organization’s AI policy before adopting any tool.
1. Planning: ask for a draft you can challenge
An AI assistant such as ChatGPT can help turn a project brief into a proposed work breakdown, identify unanswered questions, and explore alternative sequences.
Provide the constraints before asking for the plan. Request a separate list of assumptions and missing information. A plausible schedule can still rely on unrealistic durations or unavailable people.
If your organization uses Microsoft tools, check its current Planner capabilities and licensing. Microsoft retired Project for the web in August 2025 and transitioned that experience to Planner. This does not mean every Microsoft Project product was retired. Microsoft Planner FAQ.
2. Risk review: use AI to widen the review
Ask the assistant to examine project updates alongside the risk register:
Which changes might introduce a new risk? Which response lacks an owner? Which assumptions need confirmation?
Require it to distinguish documented facts from possible concerns. A language model’s suggestion is not a validated probability or a prediction of what will happen.
The useful deliverable is a better set of questions and proposed updates for the team to assess.
3. Meeting follow-through: turn discussion into action
A polished summary does not necessarily tell the team what happens next.
Meeting assistants can help prepare notes and possible actions. Microsoft Teams intelligent recap, for example, can generate notes and follow-up tasks from available transcripts, subject to licensing and configuration. Microsoft also warns that generated content can be inaccurate or incomplete. Microsoft Teams recap.
Confirm that recording, transcription, and use of the information are permitted before introducing an assistant.
Illustrative example
Imagine a project meeting where:
A supplier’s delivery date remains unconfirmed.
A scope change is discussed but not approved.
Someone offers to investigate a testing issue without agreeing to a deadline.
Ask the AI to separate decisions, proposed actions, risks, and open questions. Tell it to mark missing owners or dates as unresolved.
Then check the draft with the team. The supplier’s date must remain unconfirmed, the scope discussion must remain a proposal, and the assistant must not invent a deadline.
Ready to turn meeting notes into follow-through? Try Myappics’ free meeting workflow to organize decisions, actions, owners, and open questions using the AI assistant you already use. Review the output with your team before updating your project records.
4. Status reporting: ground the narrative in project records
Provide the reporting period, verified metrics, milestone changes, risks, and decisions needed. Ask for a concise draft that distinguishes facts from interpretation.
Check every number against its source. Missing data should appear as missing data. An explanation for a budget variance should not become fact simply because it sounds convincing.
Generative AI can produce incorrect facts and invented references with confidence. Source verification remains necessary. OpenAI’s explanation of ChatGPT accuracy limits.
5. Project knowledge: define access and authority separately
An assistant that answers questions from approved documentation has a different responsibility from one that changes records or sends messages.
Define what it may read, what it may prepare, which actions it may take, and when it must escalate. Someone must also maintain the information it relies on.
Routine work can move within agreed limits. Changes to scope, budget, commitments, or stakeholder communications need clear responsibility and appropriate checks.
How do you know whether the workflow helps?
Compare the existing process with the AI-assisted one using the same task and information.
Measure:
Total time, including preparation, review, and corrections.
Whether the output meets the requirements.
Missing information, factual errors, and invented details.
Whether the team can use the result in its normal workflow.
A second model can challenge a draft, but its agreement is not proof of correctness. Check against the project records and test the outputs that matter.
If producing the draft is faster but correcting it consumes the saved time, revise the workflow before expanding it.
Choose one task to try next
Explore the free Project Leader’s AI Field Kit from Myappics for practical workflows you can use with your existing AI assistant, from starting a project to managing delivery, taking over, or recovering one.
For practical lessons on AI in project management and everyday work, connect with Carlos "Charlie" Gil on LinkedIn.
About the author
Carlos "Charlie" Gil is the founder of Myappics, which supports PMI Memphis with digital operations and AI. He shares practical lessons from developing and reviewing AI-assisted business workflows.
