Every CFO has lived this moment.
It is 48 hours before the monthly review. The P&L looks fine. Ending cash is materially below forecast. Draft management commentary calls it “timing.”
You know it is not just timing.
You do not have 48 hours to prove it.
This is the workflow AI has been claiming to solve for 18 months. We finally tested it end-to-end this week with Anthropic’s newest model: Fable 5.
Fable 5 was released this month as a specialist reasoning model. Not the everyday workhorse. This is what you reach for when the problem is unfamiliar, the data is scattered, and the conclusion has to survive an audit committee.
Here is what you will get out of today’s letter:
The exact three-prompt workflow that turned Fable 5 from an impressive CFO analyst into a reliable one
Fixing broken month-end close with Mythos class model
How to build a report for finance
When to use Fable 5 versus Opus 4.8 in real finance work
Every CFO must build skills in Claude
If you have ever watched an AI ship a beautiful analysis that fell apart the moment you audited it, the next 8 minutes are for you.
This is the finance-grade AI workflow nobody is talking about.
Let’s dive in.
We built a demo assignment for Born2Cycle. Multi-location retail.
Real chart of accounts.
The data package was intentionally imperfect.
14 connected files. No single one explained the result.
P&L and balance sheet
13-week cash forecast
AR and AP aging
Bank transactions
Sales orders and channel performance
Customer and vendor masters
Inventory purchases and receipts
Refunds, returns and chargebacks
Payment-processor settlements
Draft management commentary
That is a real month-end close package.
Fragmented, overlapping and some fields wrong.
Commentary that reads confident but does not stand up.
We opened Claude Cowork, pointed it at the folder, and selected Fable 5 with reasoning level set to High.
Fable 5 is Anthropic’s reasoning specialist.
Extra and Max reasoning modes exist, but Max burns tokens fast and slows the response. For a 14-file finance investigation, High is the sweet spot.
Enough depth to think, fast enough to iterate.
Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.