Working With AI
Briefing, context, and review passes: the craft of getting output you can send.
Writing Down What 'Finished' Means for Client Work
"Make it professional" is a wish, not an instruction. Three tests for turning a preference about client work into a standard Claude, or anyone, can check.
When Claude's Second Draft Isn't as Good as the First
One good AI draft shows Claude can do the job. Whether it can do it again for a different client next month is a separate question. How to test it.
What's left to review when the AI checks its own work
Newer models check their own drafts before you see them. That raises the floor and makes the polish more convincing — here's the review that stays yours.
The effort dial: how much thinking does a task deserve?
Claude lets you dial how hard it thinks, from Low to Max. Which consulting tasks earn maximum effort, which are worse for it, and how to set defaults.
From transcript to deliverable: the AI-assisted research-report workflow
The four-stage workflow for turning stakeholder transcripts into a client-ready deliverable — where Claude helps, where it doesn't, and the synthesis trap.
The one-pager you should write before every AI-assisted task
The 5-section brief consultants write in 90 seconds before any AI task. Why it turns generic output into specific, and the format that compounds.
The context-window mistake most consultants make with Claude
Why your Claude conversations get worse over time, the three failure modes, and the discipline that keeps long consulting workflows sharp through the end.
Briefing, not prompting: delegating whole deliverables to Claude
How to hand Claude a whole deliverable: the five-part brief, a worked example, when to steer instead, and the review pass that keeps it client-grade.
Why your AI prompts feel generic — and how solo consultants fix it
The four reasons Claude and ChatGPT keep producing forgettable output — and the prompt structure that fixes all four for consultants.