Picking the right AI tool was never the hard part. Getting good at it is. Here’s what actually separates PMs who get value from AI from PMs who tried it for two weeks and gave up.
The Question I Get Most
Back in June, I wrote A PM’s Guide to Choosing an AI Tool — a practical framework for figuring out which of Copilot, ChatGPT, Claude, PMI Infinity, or Cursor fits your situation. It’s still the post I point people to first.
But it ended on a line I’ve been thinking about ever since: “The tool matters less than the habit.”
That post answered “which tool?” It didn’t answer the question I actually get most from other PMs, which is: “Okay, I picked one. Now what? Why does it feel like I’m getting less out of this than everyone claims I should?”
That’s what this series is about.
The Skill Isn’t the Tool
Here’s the thing nobody tells you when you start using AI at work: the tool is not the skill. Prompting is.
I use Copilot and Cursor daily at Kwik Trip. If you swapped either one out for Claude tomorrow, my output wouldn’t drop much — because the thing that actually makes the output useful isn’t which model is answering, it’s how I’m asking. Context, specificity, iteration, and knowing where to stop trusting the answer — those transfer across every tool I’ve used. The interface changes. The underlying skill doesn’t.
Most people never get past the interface. They ask a vague question, get a mediocre answer, and conclude the tool isn’t that good. Usually the tool was fine. The question wasn’t.
Four Habits That Actually Matter
These aren’t Copilot habits or Claude habits — they’re AI habits, full stop. Everything below applies whether you’re in an enterprise tool, a consumer chatbot, or a code editor.
1. Prompting Habits
The single biggest gap between people who get value from AI and people who don’t: specificity.
“Write a status report” gets you something generic you’ll have to rewrite anyway. “Write a status report for my VP audience, using these five bullet points as source material, flagging the schedule risk as the lead item, in the format we used last month” gets you something you can actually send. The second prompt takes 20 extra seconds to write. It saves 15 minutes of editing.
I’ve written about this in practice — Better Status Reports with AI and Writing RFPs with AI are both really just this habit applied to specific document types.
2. Context
AI doesn’t know your project unless you tell it. It doesn’t know your organization’s risk tolerance, your stakeholder’s communication style, or that “the vendor issue” refers to a specific vendor you’ve mentioned in six other conversations it can’t see.
The PMs who get the most out of these tools have gotten disciplined about feeding context in — pasting in the relevant background, naming the audience, stating the constraint that matters. It feels like extra work the first few times. It becomes automatic fast, and it’s the difference between an answer that’s technically correct and one that’s actually usable. Using AI to Translate Technical Jargon is a good example of context doing the heavy lifting — the same acronym means something different in a finance meeting than an engineering one, and the tool only gets that right if you tell it which room you’re in.
3. Iteration
The first response is a draft, not a deliverable. Every time.
This is the habit that took me longest to build, because it feels inefficient — why ask again when I already got an answer? But the second and third passes are where the value actually shows up: “make this shorter,” “this is too formal for this audience,” “you’re missing the risk I mentioned earlier.” Treating the first output as a conversation starter instead of a finished product is a mindset shift, not a technical one.
4. Knowing the Limits
AI is confident even when it’s wrong. That’s the part that gets people in trouble.
It can’t tell you the political weight of a phrase in your organization. It can’t tell you which stakeholder relationship is fragile right now. It doesn’t know what happened in the meeting it wasn’t in. Knowing where your judgment has to take over — and where the tool’s confidence is not the same thing as correctness — is as much a skill as prompting well. I’m still calibrating where that line sits for me, and it’s a topic I want to come back to in its own post.
What’s Next in This Series
These four habits are tool-agnostic — but they show up differently depending on what’s actually in front of you. The next three posts go deep on the tools I actually use:
- Leveling Up Copilot — M365 integration, Copilot Studio, and what changes when your AI tool lives inside an enterprise tenant with governance constraints
- Leveling Up Claude — projects, memory, and how a longer context window changes what you can hand it
- Leveling Up Cursor — codebase-aware prompting, rules files, and what “agent mode” actually means for a non-developer willing to learn it
If you only read one of the three, pick the one for the tool your organization has already approved. That’s still step one — the tool guide hasn’t changed on that point.
The Bottom Line
Two people can have the exact same AI tool and get wildly different value out of it. The difference isn’t access — it’s habits. Specific prompts. Real context. A willingness to iterate instead of accepting the first draft. And a clear sense of where the tool’s confidence ends and your judgment has to start.
None of that shows up in a product demo. All of it shows up after a few weeks of actually using the thing on real work.
Links & References
Related Monday Business Posts:
- A PM’s Guide to Choosing an AI Tool
- Better Status Reports with AI
- Using AI to Translate Technical Jargon
- Writing RFPs with AI
Coming in This Series:
- Leveling Up Copilot
- Leveling Up Claude
- Leveling Up Cursor
The tool matters less than the habit. That was true in June, and three months of using these tools daily hasn’t changed my mind — it’s just given me more specifics about what the habit actually looks like.