Using AI for Stakeholder Communication

I write a lot of stakeholder communications. Weekly status updates, executive summaries, escalation emails, quarterly business reviews. Since May, I've been using GPT-4o and Claude 3.5 Sonnet to accelerate this work. Here's what I've learned about doing it well.

The workflow

Draft with AI, edit with judgment. I give the LLM context — project status, key metrics, risks — and ask for a draft. Then I edit aggressively. The AI gives me structure and speed. I add nuance, political awareness, and tone.

Audience-specific rewriting. This is where AI shines brightest. I'll write a technical post-mortem and then ask Claude to rewrite it for a non-technical executive audience. The resulting draft captures about 80% of what I'd write manually, and I get it in 30 seconds instead of 30 minutes.

Meeting prep summaries. Before stakeholder meetings, I feed in the last three status updates and ask for a synthesized brief. It catches themes I missed and surfaces contradictions between updates.

Where I draw the line

No AI-generated emails without my review. Every stakeholder email goes through my judgment filter. AI doesn't understand organizational politics, interpersonal dynamics, or the subtle difference between "we need to discuss" and "I'd value your input."

No AI in sensitive communications. Performance issues, escalations involving specific people, contract negotiations — these require human judgment and empathy that AI cannot provide.

Transparency about AI use. I don't pretend AI-assisted writing is purely my own. When colleagues ask how I produce documentation so fast, I'm honest about using LLMs for first drafts.

The quality bar

AI makes average communicators slightly better and good communicators significantly faster. It doesn't make bad communicators good — because the editing step requires judgment that no model can substitute.

Use AI to write faster. Never use it to think less.