Aakash GuptaThe Claude Workflow Nobody at the VP Level Is Showing You
CHAPTERS
AI for VP-level work: problem decomposition beats prompt craft
Matt frames leadership-level AI as a test of how well you decompose messy problems before asking a model to solve them. If you don’t break the space into parts, the model will flatten nuance and produce confident but low-signal output.
Why most AI content fails executives: it targets IC tasks, not leadership deliverables
Aakash explains the gap between common AI-for-PMs advice (PRDs, feature analysis) and the artifacts leaders actually need (all-hands roadmaps, metrics retros, strategy narratives). The episode promises real examples from a VP using Claude for company- and board-facing work.
Case study setup: building a Q2 roadmap all-hands segment in two hours
Matt recounts waking up early to build his portion of an all-hands presentation under time pressure. The goal: communicate the roadmap to a mixed audience (GTM, marketing, sales, engineering) who weren’t in the weeds.
Start with an inventory of raw materials (before opening Claude)
Instead of immediately drafting slides, Matt first catalogs what inputs exist and which are trustworthy. He treats Claude as a transformer of materials, not a source of strategy or narrative truth.
Using Zoom transcripts to source screenshots and concrete evidence fast
Matt pulls a Zoom demo-day recording and transcript as the core raw asset, then uses Claude to extract timestamps so he can grab the right screenshots without rewatching long videos. This turns internal engineering updates into reusable evidence for broader company storytelling.
Pivoting engineering updates into strategy themes (“matrix multiplication”)
Matt combines two sources—a strategy doc with annual themes and the demo-day transcript—and asks Claude to reorganize the transcript content into the strategic categories. He describes this as a pivot/transform operation that makes the material usable for exec communication.
Slides first, talk track last: the order that improves executive storytelling
Matt builds the visual narrative first, then feeds screenshots of finished slides back into the same Claude session to generate a talk track that adds context rather than repeating the slide text. The sequencing helps maintain clarity and avoids generic narration.
Managing the ‘eager junior’ model: stop it from racing to deliverables
Matt explains a recurring failure mode: Claude behaves like an eager junior—jumping ahead to produce a full document (often in the wrong format) without asking clarifying questions. His fix is strict pacing, incremental context, and explicit commands to stop recommending next steps.
Slow-cooking clarity with abstraction: the biology metaphor method
To prevent premature solutioning, Matt starts with an abstract analogy (ecosystem/lifecycle) and withholds the business domain. This keeps the model ‘clean,’ triggers clarifying questions, and helps build a reusable mental model before mapping it onto Customer.io realities.
Layering complexity like game-night rules: foundations first, exceptions later
Matt compares good prompting to teaching a board game: explain basic rules and winning conditions first, then add exceptions gradually. Dumping nuance too early creates “indigestion” in both the model and the leader’s own thinking.
Social IQ gaps: why executive-ready writing needs political calibration and voice
Matt shows how AI can mishandle tone and “room-reading,” such as importing cute internal animal labels into executive docs. He argues leaders must own persuasion, audience empathy, and terminology choices because models lack context about baggage, timing, and credibility cues.
Decompose before you align: why AI-generated exec alignment backfires
Matt warns that using AI to “generate alignment” usually fails because executives detect slop and flattening. The outcome varies by hierarchy: seniors may get polite nods; others get ignored—either way, real alignment doesn’t happen without deep decomposition and intent.
Matt’s weekly AI stack: Claude + Slack agents for analysis, sensing, and doc audits
Matt outlines how AI shows up across his week: Claude for deep thinking and transformations, plus internal Slack agents for data analysis, conversation scanning, and strategy-document auditing. The theme is practical enablement—keeping leaders close to reality while reducing manual toil.
How to enable AI internally: controlled experimentation, secure hosting, lead-by-example
Matt explains Customer.io’s approach to tools like OpenClaw: allow experimentation in a controlled, secure way; support technical builders; and provide budget and infrastructure. Leadership adoption and feedback loops are positioned as essential for sustainable rollout.
