Last month, I watched a delivery manager spend their entire Friday afternoon copy-pasting updates from Jira into status reports, formatting slides, and chasing down team members for missing information. By 6 PM, they had a mediocre deck and a missed dinner with their family.
Sound familiar? Yeah, I thought so.
Here's the thing: while you're drowning in administrative tasks, there's a whole suite of AI tools sitting there like a life raft. ChatGPT, Microsoft Copilot, Claude—they're all ready to handle the mind-numbing stuff that makes you question why you got into tech in the first place.
After working with dozens of delivery teams experimenting with these tools, I've seen the good, the bad, and the "oh god, what have we done?" Here's what actually works—and what will have your team asking "who approved this?" in next week's retro.
Where AI Actually Helps (Spoiler: It's the Boring Stuff)
Let's start with the good news. You know all those tasks that make you die a little inside? AI loves them. It's like having an intern who never complains, never sleeps, and actually enjoys formatting documents.
DO: Let AI Be Your Administrative Assistant
Those mind-numbing status reports that eat your Friday afternoons? AI can draft them in minutes. I'm talking about pulling data from multiple sources, identifying trends, and even formatting them in that specific way your stakeholders demand (you know, the one with the exact shade of blue they insist on).
Sprint retrospective notes that nobody reads? AI can extract the patterns that actually matter. It'll spot that you've had the same blocker for three sprints straight when everyone's too polite to mention it.
Meeting summaries everyone asks for but nobody wants to write? Consider it done. And unlike Steve from engineering, AI won't editorialize about how the meeting could have been an email.
The revelation here isn't that AI is smart—it's that it's perfectly happy being boring. While you're using your brain for actual problem-solving, AI is cheerfully churning through the administrative sludge that usually clogs your calendar.
When AI Becomes Your Worst Enemy
But here's where things get spicy. For every success story, I've got a cautionary tale that'll make you clutch your laptop a little tighter.
DON'T: Let AI Make Your Decisions
I once watched a team trust an AI's sprint capacity planning that suggested they could handle 20% more story points based on their "improving velocity trend." What the AI missed? Half the team was about to be pulled into a production incident war room, and the senior developer who usually knocked out the complex tasks was onboarding a new hire. The sprint crashed and burned spectacularly.
AI can crunch numbers all day, but it can't read the room. It doesn't know that your architect just had their third kid and is running on two hours of sleep, or that the "minor" dependency you mentioned is actually a political minefield involving three teams who define "API" differently.
DON'T: Believe Everything AI Tells You
AI has this charming habit of inventing things with absolute confidence. I've seen it create dependencies between systems that have never talked to each other. It'll estimate a "simple database migration" at 2 story points when your DBA is already breaking out in hives just thinking about it.
My favorite was an AI that kept suggesting we improve deployment frequency when the real bottleneck was a change approval board that met monthly. No amount of CI/CD optimization was going to fix that human process problem.
DON'T: Let AI Run Your Team Dynamics
Here's what happens when you over-automate: Your retrospectives become AI-generated summaries that miss all the subtext. When someone says "the API integration is progressing" in a flat tone, humans hear resignation. AI hears progress.
Automated standups sound efficient until you realize you've lost your daily pulse check on team morale. AI can't see exhausted faces or pick up on the excitement when someone finally cracks that bug that's been haunting them for weeks.
Simple rule: AI can inform, but humans must decide. Think of it as a really smart calculator—useful for math, terrible at office politics. And more importantly, humans must connect.
Your Practical Playbook (No Theory, Just Action)
So here's where we stand: AI can eliminate soul-crushing busywork, but it can also confidently lead you off a cliff if you're not careful. The teams that win are the ones who know the difference. Let me show you how to be one of them.
DO: Become a Prompt Engineer (But Don't Put It on LinkedIn)
The difference between a useless AI response and a game-changing one? The quality of your prompt. It's like the difference between asking your teenager "How was school?" versus "What was the most interesting thing that happened in physics class today?"
Bad prompt: "Summarize this sprint" Better prompt: "Analyze our sprint 24 retrospective notes. Identify the top 3 impediments, their root causes, and suggest specific actions for sprint 25. Focus on technical blockers and team dynamics. Skip the fluff."
Build a library of prompts that work for your context. Share them with your team. Iterate on them like code.
DO: Set Clear Boundaries
What decisions stay human? Here's my litmus test: If you'd have to explain it in court, keep a human in charge. That means:
When is AI input helpful versus harmful?
The validation rule is simple: trust but verify everything. Especially numbers that seem too good or too bad to be true.
What Success Actually Looks Like
When you get this right, the changes are subtle but profound:
Friday afternoons become free. Not because work disappeared, but because the administrative sludge handles itself. You're thinking about next quarter's strategy instead of copy-pasting sprint metrics.
Mistakes get caught early. Not the syntax errors—we've got linters for those. I mean the "we're solving the wrong problem beautifully" mistakes that burn quarters and careers.
Your team starts acting like a team again. Developers code instead of updating tickets. PMs think strategically instead of formatting reports. Everyone focuses on delivery, not documentation.
Your Week 1 Action Plan
Monday: Recruit Your Champions Find your three champions—the experimenter who's already using AI, the overwhelmed PM drowning in reports, and the skeptical senior who'll keep you honest. Give them a simple challenge: each picks their own most painful, repetitive task and uses AI to tackle it for two weeks. Schedule a weekly check-in to share discoveries.
Wednesday: Set Your Rules Draft a one-page "When to use (and not use) AI" guide. Focus on which decisions stay human. Get team buy-in, not just compliance. This is about making everyone's life better, not adding another process.
Friday: First Check-in Host your first champion sync. What task did each person pick? What AI tool are they trying? What's their plan for week one? Set expectations: we're learning together, failures are data, successes get shared. Buy coffee. Build excitement for the experiment ahead.
The Bottom Line
Remember that delivery manager from the beginning? The one who missed dinner crafting status reports? That doesn't have to be you.
AI in delivery management isn't about replacing human judgment with algorithms. It's about freeing humans to actually use their judgment instead of drowning in busywork. It's about making it home for dinner because the reports write themselves. It's about spending your energy on the problems that actually need your brain, not the ones that just need your time.
The teams that thrive will be those that thoughtfully adopt AI for what it does best—processing information, identifying patterns, and handling repetition—while preserving the human elements that actually deliver value.
The tools are mature. The patterns are clear. The only question is: Are you ready to reclaim your evenings?
Start this week. Pick one repetitive task that makes you question your career choices. Try AI for that one thing. Your future self—and your family—will thank you.
And if it all goes wrong? Well, at least you'll have a great story for your next retrospective. Just make sure a human writes it up.