Worked example: answering routine client queries.
An accountancy practice answers the same client questions every January. Its MSP could test whether AI can draft the replies from the practice’s approved guidance in SharePoint and save the team time after review.
Illustrative client example. No real firm or client.
The client team’s review
The tool reads one approved guidance library and no client files. An accountant checks each reply before it goes to the client.
Draft for review
What do I need to send you for my self-assessment return?
Your P60, any P11D, statements of bank interest and dividends, and records of any self-employed income and expenses.
Client guidance: self-assessment checklist, section 2
“Ask the client for their P60, any P11D, statements of bank interest and dividends, and records of self-employed income and expenses.”
Sample guidance excerpt
Needs evidence
Am I going to be late?
The guidance gives the standard 31 January deadline for online returns. This client’s position is in the practice’s own records, so the tool leaves it blank and asks the team to confirm.
What would make it pay?
Measure the current process and the pilot over comparable weeks of the busy season. Include every minute the team spends checking and correcting replies.
Payback test
Time saved per reply × replies per month
Use the practice’s actual volume and staff costs to put a value on the time freed. Compare it with setup, licences and ongoing support.
Roll out if accountants can verify every reply and the result meets the practice’s payback target. Agree who owns support and can pause the tool before it goes live.
Let’s build your AI practice.
A 30-minute call with Charlie. We’ll tell you which part to start with.
Book a partner call