Every finance team already has a knowledge base. It’s just scattered: classification rules in someone’s head, the close checklist in a workbook nobody else opens, the fiscal-calendar logic rebuilt in every file.
An AI knowledge base is that same institutional knowledge, written down so an AI can search it and — critically — cite it.
Why citations are the whole game
The useful pattern is called RAG (Retrieval-Augmented Generation): the AI first retrieves relevant passages from your documents, then answers using those passages, with a pointer back to the source.
That pointer is what makes it usable in finance. A workpaper points at its support; an AI answer should too. The rule to adopt on day one: no citation, no reliance.
How to start (embarrassingly small)
- Make one folder of markdown files.
- Write down your account-classification rules — the ones in your head.
- Add your recurring reconciliation procedure, as steps.
- Add your fiscal-calendar definitions.
- Point your AI tool at the folder and ask it a question you already know the answer to. Check the citation.
Each file does double duty: new hires can read it, and agents can use it. The Skills Library is exactly this idea, published — procedures written so both humans and machines can follow them.
Accounting takeaway
Teams getting real value from AI aren’t the ones with the fanciest model. They’re the ones whose knowledge is written down. The AI Field Kit covers the rest of the vocabulary — skills, agents, and where the human reviewer stays in charge.