PM Module 04: Running the Room | Icon Alliance Academy

1

Meeting preparation with AI — arrive ready for what you know is coming

The most effective use of AI in meeting preparation isn't drafting the agenda — it's thinking through what's actually going to happen and preparing for it. Any experienced PM knows that the difficult part of a steering committee meeting isn't the prepared content. It's the question you didn't anticipate, the decision that gets more complicated than expected, or the dynamic in the room that wasn't in the pre-read. AI helps you anticipate those scenarios in advance.

Prompt — steering committee challenge preparation

Context I have a steering committee meeting tomorrow. The agenda includes presenting a scope change request for $180K integration work. I know the CFO will scrutinise any budget request. The IT Director is sceptical about the implementation partner's work. The VP Operations is generally supportive but deferential to the CFO.

Task Help me anticipate the hardest questions and challenges I'm likely to face for this scope change request, and prepare my responses.

Scope change details The original mainframe integration estimate assumed a published REST API. Discovery sessions revealed the billing system uses proprietary file-based integration with no published specification. This requires 6 additional weeks of integration work and $180K in additional partner fees. The alternative is to delay go-live by 3 months rather than incurring the cost.

Format For each likely challenge: the question or pushback I should expect, the key points in my response, and what I should avoid saying. I'll adapt the language in the room — I need the substance, not a script.

Sample AI output — challenges to prepare for

"Why wasn't this identified in the original scoping?" Key response points: the original scope was based on standard assumptions about available integration interfaces — a discovery process is specifically designed to surface this type of finding. This is a known risk of any integration project where the receiving system's interface hasn't been confirmed. What to avoid: blaming the client team for not disclosing it earlier — focus on the discovery process doing its job, not on attribution.

"$180K is a significant budget increase. What assurance do we have this is the final number?" Key response: the $180K is based on a detailed integration design estimate now that we understand the actual interface. The estimate includes a 15% contingency. The remaining integration risk is execution, not discovery — we know what we're building. What to avoid: promising it's the final number without qualification — say "this is our current best estimate based on the fully understood scope."

"What happens if we don't approve this?" Key response: the alternative is a 3-month go-live delay, which has its own costs — [quantify if possible]. The $180K is likely less expensive than a delayed go-live when operational impact is considered. What to avoid: presenting only one option — give the committee a real choice with real consequences.

Knowledge Check

During the steering committee meeting, after your prepared response to the "$180K — is this the final number?" question, the CFO asks: "What's your personal assessment — are you confident in this number?" AI helped you prepare for the factual challenge. How do you answer this question?

Correct — and this is the moment where prepared language fails and PM credibility is built or lost. When an executive asks for your personal assessment, they're asking you to step out of PM spokesperson mode and give them your actual professional judgment. That might be "Yes, I'm confident — here's why." It might be "I'm confident about the integration work itself, but I have less certainty about the timeline." Either honest answer is the right answer. The prepared estimate language was for the factual challenge. A direct question about your personal view deserves a direct personal response, including any honest uncertainty.

Option 2 is the only response that answers what was actually asked. Repeating the prepared language when asked for personal judgment signals that you're hiding behind process rather than owning a view. Deferring to the technical lead abandons PM accountability — you presented the request, the PM owns it. Reframing to process is the same evasion as repeating the prepared language. A CFO who asks for your personal assessment and gets process language has learned that you either don't have a view or won't share it — neither builds confidence. Direct, honest judgment — including honest uncertainty — is the professional response.

2

Meeting summaries and action tracking — AI compression with PM verification

Meeting summaries and action registers are high-return AI use cases for PMs — they're time-consuming to produce from notes, follow a consistent structure, and the first draft is genuinely useful. The verification requirement is non-negotiable: actions misattributed, decisions captured incorrectly, or commitments summarised inaccurately create confusion and erode the trust that clear meeting governance builds.

Prompt — post-meeting summary and action register

Context Steering committee meeting, Guidewire PolicyCenter implementation. 8 attendees including CIO, CFO, VP Operations, IT Director, implementation partner lead, and workstream leads.

Task Generate a meeting summary and action register from my meeting notes.

My meeting notes (raw) [paste raw notes — attendees, agenda items discussed, decisions made, commitments given, actions taken]

Format Summary: decisions made (each as a clear statement), items discussed but not decided (and why), next meeting date. Action register: each action as [Action description] | [Owner — specific name] | [Due date] | [Notes]. Flag any action where the owner or due date wasn't explicitly confirmed in the meeting — I need to confirm these before distributing.

The ownership verification requirement

The most common meeting summary error — AI or human — is action items with unconfirmed owners. "IT Director to confirm resource allocation by Friday" is only a valid action if the IT Director actually agreed to it in the meeting. AI flags these for you; you verify before distributing. An action register that assigns commitments people didn't make is worse than no action register — it creates disputes and erodes the governance process that keeps the project accountable.

Knowledge Check

AI generates the meeting summary and action register. One action reads: "CFO to confirm budget approval for scope change — due Friday." Your notes show the CFO said "I'll need to review this with the CEO before confirming" — she didn't commit to Friday. What do you do before distributing?

Correct — confirm before distributing, not after. The CFO has a specific process (CEO review) that affects the timeline. Assigning Friday as a due date she didn't commit to puts her in a position of either failing a deadline she didn't agree to or having to correct the record publicly. Neither is conducive to the governance relationship you're building. A 5-minute call before the summary goes out — "I want to make sure I've captured this accurately — can we agree a realistic date given the CEO review?" — is the professional approach. The action register is only useful as a governance tool if its contents are accurate.

Option 3 is the only complete response. TBD is better than a wrong date but still distributes an action without confirmed timing — leaving the urgency ambiguous. Following up separately after distributing means the incorrect or incomplete action is already on record. The right sequence is: confirm first, then distribute with accurate information. A brief pre-distribution call to the CFO to confirm the date takes 5 minutes and produces an action register that actually reflects what was agreed. That's the governance standard the PM should hold.

3

Difficult conversation preparation — AI as rehearsal partner

Difficult conversations in project delivery — telling a sponsor the project is behind, addressing a team member's underperformance, managing a workstream lead who is consistently missing commitments — benefit from preparation. AI helps with the preparation: structuring what you want to say, anticipating responses, and thinking through the conversation before you're in it. What happens in the actual conversation is entirely yours.

Prompt — difficult conversation preparation

Situation I need to have a difficult conversation with a client-side workstream lead who has missed the last three action commitments without advance notice — he just doesn't deliver by the agreed date and then explains why afterward. This is now affecting the testing schedule. He's well-liked by the business sponsor and has strong organisational standing.

Task Help me prepare for this conversation. I want to address the pattern directly without damaging the relationship or the programme. I need to come out of it with a different working arrangement — not just an apology and another commitment.

What I know He's genuinely busy — not malicious. He takes on commitments without saying no even when he's overloaded. The programme impact is real and I need a different behaviour, not just acknowledgement. His relationship with the business sponsor means I need to handle this without it becoming a political issue.

Format Opening statement (factual, non-accusatory), what I'm asking for specifically (a changed working arrangement, not just an apology), how to handle the likely response ("I've been very busy"), and what I want to agree by the end of the conversation. This is a framework — I'll adapt in the room.

What AI preparation gives you — and what it doesn't

AI preparation for a difficult conversation gives you: a structured opening that doesn't start with accusation, language for the specific ask, anticipated responses and how to address them, and a clear picture of what outcome you need. What it doesn't give you: the ability to read how the conversation is actually going and adapt in real time, the judgment to know when to push and when to let something breathe, or the interpersonal read on how the other person is receiving what you're saying. AI gives you a map. The territory is the actual conversation, which will go differently than the map in at least one significant way. The preparation makes you more effective in the room; it doesn't substitute for what happens there.

Knowledge Check

Halfway through the difficult conversation, the workstream lead becomes defensive and says "I feel like I'm being ambushed — I wasn't told this conversation was going to be a formal performance discussion." Your AI preparation covered this as a possible response. What's the right move in this moment?

Correct — and this is the real-time judgment that no amount of AI preparation substitutes for. Defensive reactions in difficult conversations are information: he's feeling threatened, not engaged. The right move is to reduce the threat perception — acknowledge the feeling, clarify your intent — while holding the ground on why you're there and what you need. Continuing rigidly with the prepared framework when the person has become defensive usually makes things worse. Conceding entirely and rescheduling loses the momentum and the programme impact doesn't go away. The adaptive middle — acknowledge, clarify, redirect to the specific outcome — is what the preparation made you ready to improvise. That improvisation is yours.

Option 3 describes the adaptive professional response. Continuing the prepared framework without acknowledging a clear defensive reaction escalates rather than resolves the tension. Rescheduling concedes the ground entirely and allows the pattern to continue another week while the testing schedule slips further. Validating his workload constraints fully before returning to impact risks never returning to impact — a conversation about his constraints is useful but only if it leads to a specific changed arrangement, not as a substitute for it. The real-time judgment — acknowledge the feeling, hold the substance, redirect to the outcome — is what effective conversation management looks like when the preparation meets the actual room.

4

Go/no-go decision discipline — the PM's most consequential call

Go/no-go decisions are where PM professional judgment is most consequential and most visible. AI can help you structure the go/no-go framework, draft the assessment, and think through the decision criteria. The recommendation is yours — and the professional courage to make a "no-go" recommendation when the data supports it, even under significant schedule and political pressure, is what separates delivery-focused PMs from schedule-focused ones.

✅

What AI can do for go/no-go

Generate the exit criteria framework for a phase gate or UAT exit. Structure the go/no-go assessment document. Draft the options analysis if the recommendation is conditional. Help you think through the consequences of each decision. Provide a structured format for presenting the recommendation.

🧠

What the PM must provide

The honest assessment of whether the exit criteria are genuinely met — not "close enough given the pressure." The professional courage to recommend no-go when the data supports it. The judgment about which outstanding issues are genuinely manageable in production versus which ones will cause failures. The recommendation that the PM is prepared to stand behind if it turns out to be wrong.

⚠️

The false go — the most common failure

The go/no-go decision made under schedule pressure where known risks are rationalised rather than addressed. "We'll fix it after go-live" is the most expensive phrase in insurance IT delivery. AI-assisted go/no-go documentation can make a false go look more professional — the assessment looks thorough even if the recommendation doesn't reflect what the PM actually knows. The discipline is making the recommendation match the assessment, not the schedule.

💼

The PM's professional accountability

A PM who recommends go when they know the criteria aren't met — because the sponsor wants go-live and the political pressure is significant — is making a decision they cannot take back. When the production issues materialise, the PM's recommendation is on record. Professional PMs recommend what the data supports and explain clearly what the risks of proceeding are. The decision belongs to the governance level; the recommendation belongs to the PM.

5

Module summary

✅

Prepare for what's actually coming

AI meeting preparation is most valuable for anticipating challenges, not just drafting agendas. Identify the hardest questions and pressure points, prepare the substance of your responses. In the room, adapt — the preparation gives you the material; the judgment is real-time.

✅

Verify before distributing

Meeting summaries and action registers require owner and date confirmation before distribution. An action no one committed to is worse than no action. Contact owners before sending, not after. The action register is a governance tool — its value depends entirely on its accuracy.

✅

Framework for difficult conversations, not scripts

AI gives you structure, anticipated responses, and the specific ask. The conversation itself requires reading the room and adapting. Acknowledge defensive reactions without conceding the substance. Hold the outcome while adjusting the approach. That judgment is yours.

✅

Recommendations match the data

Go/no-go recommendations reflect what the assessment actually shows, not the schedule pressure. "We'll fix it after go-live" produces predictable failures. The PM's job is to make the recommendation the data supports and explain the risks of proceeding clearly. The governance level makes the decision; the PM owns the recommendation.

One module left

Module 05 — Your AI-Augmented PM Practice — brings the full pathway together. Daily habits, market positioning for senior PM engagements, and a personal readiness assessment across all five modules.

✓

Module 04 Complete

Running the Room is done. Continue to Module 05: Your AI-Augmented PM Practice.

Back to Pathway Continue to Module 05 →