Explain with AI
Part of the Close plan: 200 explanations a month, and 25 during a free trial.
Explain with AI writes a short explanation of one thing you're looking at, in plain English, with what to check in GP. It's there for the moments when the numbers are right in front of you but the reason isn't. It never changes anything in GP, and nothing runs unless someone chooses the button.
What it explains
- A flux variance: why a GL account moved between periods, from the change by vendor, customer or reference and the largest entries. Use as the explanation copies it into the account's explanation for you to edit.
- AP and AR reconciling items, on the Reconciliations page: what each kind of item on the tie-out means for your company, and what's left unexplained.
- A check's findings, on the check's page: what went wrong, why GP ends up that way, and how to resolve it.
- A close, on the Close page: a short summary for a controller of what's done, what's open and what needs attention.
Turn it on
It's off until an approver turns it on. In Settings, under Explain with AI, read what's sent, check Turn on Explain with AI, and choose Save. Unchecking it turns it off for everyone.
What's sent, and what isn't
Only what's needed for the one thing being explained leaves your server. Ledger Doctor builds it on your server from the same numbers the page shows:
- Sent: amounts, dates, account numbers and descriptions, document numbers, document types, GP sources, and the check's own description.
- Replaced before sending: vendor, customer, employee and user names and IDs, and free-text fields such as references, descriptions and batch names. Each becomes a code like VENDOR-1 or TEXT-3. The codes are turned back into the real names on your server when the answer arrives, so the AI never sees them.
- Never sent: your company's name, the server's name, your license key, or anything about other accounts, checks or periods.
It goes from your server to myledgerdoctor.com, which counts it against your allowance and passes it to Anthropic's Claude. Anthropic doesn't train its models on it. myledgerdoctor.com keeps only the count and its size, never what was sent or the answer. The explanation itself is kept on your server, so opening it again doesn't use another one.
Amounts are checked
Every amount in an answer has to match a number in what was sent, to the cent. If an answer mentions any other amount, it's asked again once; if it still doesn't match, it isn't shown and isn't counted. Still, read it as a starting point: the AI can misjudge why something happened even when every number is right.
Your allowance
The Close plan includes 200 explanations a month, and a free trial 25. The button shows how many are left, and the count starts again on the 1st. Ask again gets a fresh explanation and uses one. Essentials doesn't include Explain with AI; upgrading adds it.
How do I…?
Choose How do I…? at the top or the foot of any page and ask how to do something in Ledger Doctor, for example "How do I turn on HTTPS?". The answer comes from this help center only, with links to the pages it used; if the help center doesn't cover it, it says so. Only your question is sent to myledgerdoctor.com, nothing from GP. Every plan has 50 answers a month, counted apart from Explain with AI, and they need AI turned on in Settings. Answers are written by AI, so check the page it links to before you change anything.