The email sits in your drafts folder because you don’t know what to write — or you know what you want to say but can’t figure out how to say it without blowing up the relationship.
Every PM has a version of this problem.
You need to push back on a vendor who isn’t delivering. You need to tell a manager that their team’s lack of availability is putting your project in the red. You need to escalate without making it personal. You need to be direct without being aggressive, firm without being harsh, clear without being brutal.
And so the draft sits. You open it, write a sentence, delete it, close the tab.
This is where AI has become genuinely useful in my work — not as a replacement for judgment, but as a drafting partner that helps me find the right words when I’m stuck.
The Problem AI Actually Solves
Most advice about difficult conversations focuses on the conversation itself — how to read the room, how to manage emotions, how to negotiate in real time. That’s all valuable. But the harder problem for a lot of PMs is getting started.
When you’re staring at a blank email to a vendor who has been ignoring your team for three weeks, you’re not thinking about body language. You’re thinking: how do I say this in a way that gets results without making things worse?
AI is good at that specific problem. Give it the right context and it will draft something with the appropriate tone — professional, direct, outcome-focused — that you can then edit for what you know about the person and the situation.
Three Real Examples
Vendor Not Responding — Product Evaluation at Risk
We were evaluating a vendor’s tool to determine if it would work for our environment. To complete the evaluation, we needed their team to help us configure it. The people assigned to work with us had gone quiet — emails unanswered, calls unreturned. The evaluation was stalling. I needed to escalate to the account manager to get things moving.
I gave AI the background: what we needed, what we’d asked for, how long we’d been waiting, and what I needed to happen. I told it the tone I wanted — firm but professional, focused on getting a response and a timeline, not on expressing frustration.
The draft it produced was better than what I would have written on my own. It was direct without being hostile. It made clear that we needed action, gave them a specific ask, and left the door open for a productive response. I edited it for a few details I knew about the relationship, sent it to the account manager, and got a response the same day.
Vendor Causing Project Delays
A second vendor situation was more serious — they were slow to fix blocking issues, and that was causing actual project delays. This wasn’t just an evaluation stalling; unresolved blockers were affecting our timeline and our stakeholders were noticing.
The email I needed to write was harder because the stakes were higher and the tone had to thread a narrower needle. Too soft and nothing changes. Too aggressive and we damage a relationship we still need.
I used AI to draft the escalation. I gave it the full picture: what the delays were, what the impact was, what I needed from them and by when. The AI structured it clearly — here’s what’s happening, here’s the impact, here’s what we need, here’s when we need it. It read as serious without being accusatory. It gave the vendor a clear path to fix the situation.
Telling a Manager Their Team Is the Problem
This one is uncomfortable in a different way. When a project goes yellow or red because a team can’t prioritize the work, someone has to say it — and that someone is usually the PM.
The challenge is that you’re essentially telling a manager that their decisions are creating a problem for your project, without making it sound like an attack, and without understating what’s actually happening.
I gave AI the situation: project status, what work was blocked, what the team’s availability looked like, what I needed from the manager. I asked for a tone that was factual and solution-focused — here’s where we are, here’s why, here’s what needs to change.
The draft framed the situation around the project impact rather than the manager’s decisions, which made it easier to have the follow-up conversation. It gave me a starting point I could work from instead of a blank page.
Where AI Falls Short
I want to be honest about this because I think people either oversell AI for interpersonal situations or dismiss it entirely.
AI doesn’t know the person you’re emailing. It doesn’t know if your vendor contact responds better to data or relationship-building. It doesn’t know if your manager takes direct feedback well or gets defensive. It can’t read the room because it has never been in the room.
It also can’t anticipate how someone will actually react based on personality and history. It can give you a list of potential objections to prepare for — and that’s useful — but the list is generic. You have to apply your own knowledge of the person to figure out which objections are actually likely.
And it absolutely cannot replace your judgment in a live conversation. If you’re in a meeting and things go sideways, AI isn’t going to help you in real time.
How to Use It Well
The prompt matters more than most people realize. A vague prompt gets a generic draft. A specific prompt gets something actually useful.
When I use AI for a difficult stakeholder email, I give it:
- The background — who the person is, what the relationship is, what’s happened so far
- The specific situation — what went wrong, what the impact is, how long it’s been going on
- The action I need — exactly what I want them to do and by when
- The tone — firm, professional, solution-focused, not accusatory
Then I edit the output for what I know about the person. The AI gets the structure and tone right. I make it accurate to the specific situation and relationship.
Think of it as a first draft from a colleague who writes well but doesn’t know your stakeholders. Their draft is better than a blank page. Your edits make it right for the actual situation.
The Bottom Line
Difficult emails sit in draft folders because the blank page problem is real. You know the situation. You know what you need. You don’t know how to say it.
AI solves that specific problem well. It gives you a starting point with the right tone, surfaces questions you might not have considered, and gets you unstuck.
It doesn’t replace your judgment about the person, the relationship, or the room. But it gets you from zero to something you can work with — which is often the hardest part.