Smarter Projects, Sharper Decisions: How I Use AI to Rethink Delivery
This article was originally posted on LinkedIn on July 26, 2025
When people hear “AI in project management,” their minds often jump to automated scheduling or virtual assistants. Those tools have their place, but they barely scratch the surface of what artificial intelligence can do to transform how we lead complex projects.
As a senior project manager and doctoral student focused on large-scale IT delivery, I’ve been researching and prototyping ways to use AI that go far beyond time-saving. I’m exploring how it can reshape how I assess risk, track performance, and support real-time decision-making.
This isn’t about replacing the human role. It’s about augmenting it. Here’s the research path, strategic mindset, and experimental tooling that shape my approach.
1. AI as a Decision Support Partner
Most project dashboards tell you what already happened. My research explores how AI can help anticipate what’s about to happen. I’ve been experimenting with lightweight GPT-based assistants that interpret project data (status updates, time logs, resource allocations) and surface:
- Early indicators of scope creep or schedule slippage
- Gaps in communication patterns across team updates
- Risk signals based on language sentiment in weekly status reports
These tools aren’t making decisions, they help PMs ask better questions. They’re like having an always-on project coordinator who never gets tired of pattern recognition.
2. Natural Language Interfaces for Project Queries
One line of exploration has involved natural language query interfaces. I’ve prototyped a chatbot that connects to structured project data and can answer:
- “Which tasks are behind schedule this week?”
- “Who is overallocated in Sprint 3?”
- “Show me variance in time estimates for the top 5 delayed features.”
Although not yet implemented in production, this prototype helped shape my understanding of how AI can reduce the friction in project analytics and improve access to insight across non-technical roles.
3. AI for Meeting Summaries and Action Extraction
Manual meeting summaries and follow-up lists are time-consuming. I’ve tested AI transcription paired with prompt-based summarization to extract:
- Key decisions
- Owner/action pairs
- Implicit blockers (e.g., “We can’t proceed until X is resolved”)
While imperfect, this process reduces administrative effort and increases transparency — and informs my ongoing investigation into language model reliability for task tracking.
4. AI-Assisted Risk Identification
I’ve also explored using LLMs to process qualitative project notes and generate structured risk input. Given a set of updates, the assistant can:
- Suggest overlooked risk categories
- Draft mitigation strategies informed by prior case patterns
- Identify inconsistencies between task progress and stated risks
This approach is still in an early research phase, but it offers promising angles for reducing blind spots and cognitive bias in risk planning.
5. A Research-Driven, Iterative Approach
The most important lesson in applying AI to project management? Treat the tools as iterative research outputs, not plug-and-play products.
My primary assistant is a GPT-based prototype tuned with project management context, PMBOK principles, and my own documented project experiences. I’m not chasing full automation. I’m exploring leverage, insight, and decision support, while preserving governance and security boundaries.
Final Thoughts
Exploring AI in project management hasn’t made me obsolete, it’s sharpened my thinking. It’s helped me shift from tracking the work to truly understanding the dynamics behind it.
In a profession where PMs are expected to deliver more with less, AI isn’t a buzzword, it’s an emerging discipline. And for those of us willing to explore, it may be the key to managing smarter, not just faster.
If you’re starting your own AI journey in project leadership, begin with the friction points. What question do you answer over and over? What task makes you feel like a bot? Start there and let the research evolve.
AI Assistance Disclosure: Portions of this post were developed using GPT-4 for editorial refinement. All insights and experimental frameworks reflect real-world research and prototyping within my professional practice.