At the weekly cash meeting the CFO asks a plain question: which customers need a person this week, and who is on each one? The answer arrives on Thursday. It has been assembled from the collections workspace in Dynamics 365 Finance, an export of open items, the unapplied cash list and a few emails from sales. By then half the week has gone.
That gap, between what the ERP records and what the team needs to act on, is where AI in order-to-cash earns its keep. Dynamics 365 Finance now includes useful AI features. Most of them work one record at a time. The CFO's question spans all of them.
What Dynamics 365 Finance now offers
Two features matter most for order-to-cash. Both are described on Microsoft Learn pages updated on 1 September 2026.
Summary by Copilot for collections. Microsoft describes it as an "AI-generated summary of a customer's overdue invoices, payment history, and remaining credit", with "an AI-generated draft of a reminder email". The summary draws on "payment history for the past year, outstanding debt amount, and invoices for the last six months". For a collector preparing a call, it means a paragraph to read and check instead of moving between screens.
Account reconciliation. Available from version 10.0.42, it reconciles the general ledger with "accounts payable, accounts receivable, tax, and bank subledgers". Microsoft says the Copilot-powered agent, where enabled, "suggests the most likely action to take", and it fills in AI-generated fields for a suggested action, an analysis and a justification. For receivables, that gives the accountant a first explanation of why the subledger and the ledger disagree.
Check what your version, licensing and configuration actually enable before planning around either. Both are assistive. The collector still decides what to send, and the accountant still decides what to post.
What a layer over Dynamics adds
The built-in features answer questions about one customer or one account. A finance leader manages exceptions across the whole book.
| The question | Where it is best answered |
|---|---|
| What does this customer owe, and how have they paid? | Inside Dynamics: Summary by Copilot on the customer |
| Draft a reminder for this customer | Inside Dynamics: the Copilot draft, reviewed by the collector |
| Why does receivables not agree to the ledger this month? | Inside Dynamics: account reconciliation, with the agent's suggestion |
| Which customers, across the whole book, need a person this week? | Across records: a ranked exception view |
| Who owns each dispute and deduction, and how long has it waited? | Across records: ownership and aging by item |
| What happened to last month's exceptions? | Across records: a history of actions and outcomes |
The three lower rows are what a layer over Dynamics adds. It reads from Dynamics and leaves the posting there. Newly late accounts, broken promises, aging disputes, unapplied receipts and open deductions from every customer land in one list, each with one owner and a clock. The layer can also bring in what Dynamics does not hold, such as dispute emails or notes from the account manager.
We covered the ranking logic in Collections Prioritization: Who to Call First, and Why the Aging Report Gets It Wrong and the receipts problem in Unapplied Cash and Short Payments: The Quiet Drag on DSO. The layer is where those methods run every day.
Where to start with AI in order-to-cash
1. Test what you already license. Turn on the Copilot collections summary for a few collectors and the reconciliation agent for the receivables control account. After a month, ask the collectors whether the summaries were accurate and whether they used them.
2. Pick one exception queue. Disputes, unapplied cash or deductions: choose the one with the most value tied up and the vaguest ownership today. Resist building the whole order-to-cash view at once.
3. Build the cross-customer view for that queue, read-only. Pull the data from Dynamics, assign owners, show the age of each item and keep a record of what was done. Nothing writes back to the ledger at this stage.
4. Keep a person on anything that leaves the team. Customer emails are drafted and then sent by a collector. Credit holds, write-offs and accepted deductions go through the approval limits you already have.
5. Measure before you widen. Track the age of the chosen queue, the share of promises kept and how long disputes take to resolve, alongside DSO. APQC Open Standards Benchmarking (cross-industry median, all companies) gives a median DSO of 38.0 days from 13,195 companies, but your own trend is the more useful comparison.
What AI in order-to-cash will not fix
It will not fix customer master data. Wrong payment terms, outdated bill-to contacts and missing purchase order references produce disputes that no summary prevents. A Copilot summary built on wrong terms is fluent and wrong.
It will not make commercial decisions. Whether to put a large customer on credit hold, accept a deduction or agree a payment plan depends on the relationship, the margin and what sales knows. AI can lay out the facts. The decision and the accountability stay with named people.
Frequently asked questions
Does Dynamics 365 Finance have AI for collections?
Yes. Summary by Copilot produces an AI-generated summary of a customer's overdue invoices, payment history and remaining credit, and an AI-generated draft of a reminder email, according to Microsoft Learn documentation updated in September 2026.
What does the account reconciliation agent in Dynamics 365 Finance do?
It helps reconcile the general ledger with the accounts payable, accounts receivable, tax and bank subledgers. With the Copilot-powered agent switched on, each difference gets a proposed action, an analysis and a justification for a person to review.
Where should a CFO start with AI in order-to-cash?
Start by testing the AI features you already license, then pick one exception queue, such as disputes or unapplied cash, and build a read-only cross-customer view with owners and aging for that queue before widening.
Do we need a separate tool on top of Dynamics 365?
Not always. If per-customer summaries and the reconciliation agent cover your needs, start there; a layer adds value when you need one view of exceptions across all customers, with an owner and an age on each, and data from outside Dynamics.
Can AI send collection emails automatically?
It can, but it should not at the start. Have AI draft the email, and have a collector review and send it, so that tone, accuracy and any open dispute are checked by someone accountable for the customer relationship.
Where CFOLogic fits
Order-to-cash on Microsoft Dynamics 365 is one of the approaches we work on with CFOs: a read-only layer over Dynamics listing exceptions across customers, each with a named owner and an age, and a person approving anything sent outside finance. We would scope a first version as a proof of concept on a single queue. For the same pattern applied to payables, with a fictional company, see the purchase-to-pay demo.