Skip to main content
One plain-language ask, no file references or account codes. The client works the data in Control unattended and comes back with the variance story and a self-contained interactive report.
The recording has narration, and the figures, entities, and account names in it are illustrative. Nothing in the video is a target or a benchmark.

​
What happens in the recording

  1. The user asks a connected AI client for a budget-versus-actuals analysis of last quarter.
  2. The client finds the workspace budget on its own and pulls the quarter’s actuals plus the prior quarter.
  3. It works unattended — sixteen tool calls over about ten minutes: FX conversion, cost-center allocations, duplicate screening, and transaction-level drilldowns of the outliers.
  4. The answer leads with the story: where profit landed against plan, and whether the miss is revenue or cost — with a variance chart per statement line.
  5. The deliverable is one interactive file: KPIs, an expandable statement, cost centers, a vendor drilldown of the over-budget account, and data flags — ready to download and share.

​
Try it in your workspace

  1. Connect an AI client with MCP — read access is enough.
  2. Ask the way you would ask a colleague: “how did last quarter go against budget?” If the workspace has several budgets or reporting bases, name the ones to use before the analysis starts.
  3. Ask which variances are timing rather than performance, and to drill the real outliers down to vendors and transactions.
  4. Ask for the result as one self-contained report you can share, flags included.

​
Read the flags before the numbers

The caveats are part of the analysis: an approximate allocation, a vendor view built from actuals only, months that look like churn but are probably unbooked invoices. The client should surface them with evidence in the report — a report without flags reads cleaner and is worth less.
Example prompts: “Do a budget-versus-actuals analysis for last quarter and package it as one interactive report I can share.” “Which variances are timing rather than performance?” “Take the one account that is clearly over budget down to vendors and transactions.”

​
Troubleshooting

​
The vendor drilldown is empty or actuals-only

A vendor-level budget comparison needs a counterparty split in the budget tree. If the budget does not carry one, the client can still build the view from actuals — it should just say so.

​
The analysis stops at the statement lines

Ask the client to take the largest deltas down to transactions and name the drivers — a specific vendor, a single event, one entity’s spike — and to separate one-offs from run-rate changes.

​
Next steps

Adjust a budget

Turn the findings into a reforecast, in reviewable layers.

Budget versus actual

Run the comparison on a reconciled basis with a packaged skill.

Monthly management report

Make the quarter-end package a monthly routine.

Create a budget from scratch

Seed next year’s plan from the actuals you just reviewed.