If you're a CFO who's been hearing about AI for months and still haven't found your entry point, here's the most honest advice I can give you: stop waiting for the perfect use case to reveal itself. Just ask Claude to teach you Claude.
That's exactly what I did. And it took twenty minutes.
The Starting Point: Total Beginner
I want to be upfront about where I was coming from. I've spent a lot of time recently with Claude + Excel — that's become familiar territory. But Claude Cowork was completely foreign to me. I'd never coded a thing in my life. I'm not even particularly advanced in Excel. And I had no idea what Cowork could actually do for a finance leader.
What I did know was that I wanted to find out. So instead of Googling tutorials or watching YouTube videos, I did the thing that's becoming second nature: I asked Claude to build me a guide.
The result was a complete CFO's implementation guide for Claude Cowork — covering setup, security, ten production-ready finance workflows, and a full scaling playbook — and I didn't write a single word of it myself. Here's exactly how it happened.
Step 1: Get the Outline Right Before You Build Anything
The first instinct most people have with AI is to ask for everything at once. I've learned that's usually a mistake. So I started with the structure, not the content.
That last sentence — "align on that outline first" — is the most important part. Getting the structure right before Claude goes off and builds everything saves enormous time. It takes thirty seconds to redirect a table of contents. It takes twenty minutes to redirect a fully built guide.
We went back and forth a few times on the structure — adjusting the workflow selection, tightening the section order, making sure the four parts felt right for a CFO audience. The guide landed on: Foundation, Implementation, the 10 Workflows, and Scaling. Only once the outline felt right did I say: go build it.
I went away, took a shower, and came back to a finished v1.
Step 2: Push It Past "Good Enough for a Technical Audience"
The first version was solid. Comprehensive. Well-structured. But reading through it, something felt off — it assumed too much familiarity with the tools. A first-time AI user would get three steps in and feel lost.
This is the prompt that most people skip. They take the first version, decide it's good enough, and move on. But the gap between "technically correct" and "actually useful for a real person trying this for the first time" is enormous — and it only costs one more prompt to close it.
Off it went and rebuilt the workflows with proper first-timer scaffolding. Each workflow now follows the same five-step pattern: gather your files, open Cowork, paste the prompt, click run, review the output. More back and forth to refine individual sections. Each iteration tighter than the last.
Step 3: Make Claude QA Its Own Work
This is the step I didn't expect to matter as much as it did.
Claude went through the guide, found errors in some of the step-by-step instructions, fixed them, and came back with a cleaner version. Not perfect — I still reviewed it myself — but materially better than what I would have caught in a single read-through.
Asking an LLM to QA its own output sounds like asking someone to proofread their own work. And yes, it has limits. But it catches more than you'd think — broken logical sequences, steps that assume prior knowledge they haven't established, instructions that are technically correct but practically confusing for a first-timer. It's worth the extra prompt every time.
Twenty minutes after I started, I had something genuinely shareable.
The entry point to AI isn't a killer use case. It's the willingness to treat "I don't know where to start" as a prompt rather than a reason to wait.
Three Things I Learned
-
01
Don't trust the first version of anything an LLM gives you. There are going to be mistakes. Gaps. Assumptions that don't hold for your audience. The first version is a starting point, not a deliverable. Treat it like a first draft from a junior analyst — good bones, needs refinement.
-
02
The more context you give, the more tailored the output. Telling Claude you're building a guide "for CFOs who are first-timers in AI" produces a fundamentally different result than "build me a Cowork guide." Context isn't just helpful — it's the difference between generic and genuinely useful.
-
03
Don't aim for perfection. Aim for 70%. A complete, coherent, shareable CFO guide to Claude Cowork in twenty minutes is a massive win. It's not the final word on the subject. But it exists, it works, and it was built in the time it used to take to format a spreadsheet.
Monthly close · Board packs · Expense compliance · Cash forecasting · Variance analysis
