AI data analysis for FileMaker: ask questions, get answers

Most FileMaker files hold the answers to questions nobody has time to build a report for. AI data analysis closes that gap: you ask in plain English, an AI assistant such as Claude looks the data up in your file, and it replies with the answer and the figures behind it. Ailimu is the connection that lets it do that against live FileMaker data.

What it looks like

You ask the way you would ask a colleague:

"Which trade customers have overdue invoices, and how much do they owe?"
"How did March sales compare with February, by product line?"
"Which customers ordered every month last year but have not ordered since?"
"What are our ten biggest customers this quarter, and how many of them also appear in last quarter's ten?"
"Which orders are still open more than 30 days after they were placed?"

Claude works out which tables and fields answer the question, looks them up, does the arithmetic and writes the reply. You can follow up in the same conversation: "now only the ones in Scotland", "put that in a table".

How it works

Ailimu runs on your own server and sits between Claude and your FileMaker file. It gives Claude a short list of tools, such as "find records in Invoices" or "run this script", and Claude calls them as needed. Three things follow from that:

The data stays live. Claude reads the file as it is now. There is no export step and no spreadsheet to go stale.
No second copy. Records are fetched to answer a question, not loaded into another system.
You set the limits. Ailimu is read-only until you turn editing on for a table, and Claude can only see the fields you chose. See Give your team AI access to FileMaker, safely.

Every lookup is recorded in Activity in the wizard, so you can see what was asked of your data.

What it is good for

One-off questions that would otherwise need a new find, a new layout or a new report.
Comparisons and summaries across a period, a customer group or a product line.
Spotting things to look at, such as customers who stopped ordering or invoices past due.
Explaining a number you already have: "why did this month's total drop?"

What it is not

Not a replacement for the reports you run every day. A tuned FileMaker report is faster and gives the same figure every time.
Not always right. An AI assistant can misread a field or choose the wrong table. Check the first answers against a report you trust. Clear field names help, because Claude relies on them to pick the right one.
Only as good as the data. If dates are stored as text or half the customers have no region, the answer will reflect that.

Getting started

Connect Claude to your FileMaker data. About ten minutes, read-only from the start.
Ask a question you already know the answer to and compare.
Then ask one you do not.

Prefer the terminal? Use Claude Code with FileMaker.