Published: 2 November 2026
Audience
- Finance Team
Objective
This article explains how to run a short, friendly billing data quality check with Claude or ChatGPT. It shows what is already in good shape and what to fill in, so the deeper finance recipes give the best answers.
Prerequisites
The MCP connector is active on your organization.
An Admin has enabled the MCP connector for your organization.
Your AI assistant is connected.
Your AI assistant (Claude or ChatGPT) should be connected to PSOhub.
See How to Connect Claude to PSOhub or How to Connect ChatGPT to PSOhub.
Steps
- Copy the prompt exactly as written:
-
Run a setup check on our billing data: is it ready for deeper analysis? This is not an audit. Every organisation has gaps when they start, and that is normal.
Look at: whether sent invoices are actually marked as sent, the share of hours that carry a cost rate, invoices that still carry the number Draft, and whether stored base currency amounts are filled on invoices, hours and contract lines.
Start with what is already in good shape. Then give at most four improvements with the biggest effect, in one table: what to fix or fill in, roughly how many records, and which question gets a better answer once it is done.
Say which scope you counted. Do not list individual records, but offer three examples of any improvement. Keep it positive and practical. - Paste the prompt into Claude or ChatGPT and send it.
Result
The answer starts with what is already in good shape in your billing data. It then shows up to four improvements in one table, with what to fill in, roughly how many records it affects, and which question gets a better answer once it is done.
💡 Tip: Ask a follow-up question, for example:
- Which of these gives the biggest result for the least work?
- Show three examples of the first improvement.
- Run the check again (after you have fixed something).