Flux analysis compares your books period over period, flags the accounts that moved more than you'd expect, and explains what drove each move. In Puzzle you can run it three ways: as an AI task in Close, with Spotlight on the Cash Activity report, or from Claude, ChatGPT or another AI assistant through the Puzzle MCP.
Run a flux analysis in Close
The Variance (flux) analysis task compares the selected month with the prior month, flags every account that moved more than 5% or $1,000, and writes commentary you can use in a monthly close report.
Go to Close under AI Suite in the left sidebar.
Click Add, find Variance (flux) analysis in the templates, and add it to your checklist. You can edit the instructions, for example to change the thresholds.
Click Run. The agent pulls both periods from the ledger, computes the dollar and percentage change for every account, and looks at the transactions behind each flagged account.
Click Review when the task shows Ready. Each flagged account lists the change, the transactions that drove it, and a High, Medium or Low confidence for the explanation.
To run it every month, click Add schedule on the task and choose Monthly.
Close is available to accountants and firms that manage the company's books. If you don't see it, open Cowork under AI Suite: the same Variance (flux) analysis template is there.
The agent explains what it finds. It doesn't change your books without your approval.
Spot changes on the Cash Activity report
On the Dashboard, open the Cash Activity report and click Spotlight. Spotlight highlights fluctuations from one period to the next, so you can see where cash flow changed at a glance.
Run a flux analysis from your AI assistant
With the Puzzle MCP connected, ask your assistant something like "Run a flux analysis on our income statement for September vs August. Which accounts moved the most, and what drove them?"
The assistant compares the two periods for the income statement, balance sheet or trial balance, flags material changes, then pulls the ledger entries behind each flagged account to explain them.