Accounts receivable automation, built as an AI agent on your own books.

It means software does the repeated steps of getting paid by your customers. It sends invoices, reminds customers before and after the due date, matches each incoming payment to the invoices it pays, and reports what is overdue. People still decide who gets credit, how a dispute ends and what is written off. Accounts receivable, or AR, is the money customers owe you and the work of collecting it.

Work in order. Send every invoice by email from your accounting system, so it can be tracked. Turn on the reminders your system already has. Feed your bank into the books so payments arrive as data. Write matching rules for payments that carry an invoice number. Then add a person, or an agent, for what is left: payments with no reference, short payments, disputes and the accounts that need a phone call.

Cash application is matching each customer payment to the invoices it pays and closing them in the ledger. It is slow when the remittance, the note saying what a payment covers, arrives apart from the money, by email or PDF. Automating it means reading the bank line and the remittance, finding the open invoices by number, name and amount, and leaving unclear matches for a person.

Three cover most of it. Days sales outstanding, or DSO, is roughly how many days of sales are still unpaid. The aging report shows what is current, 1 to 30, 31 to 60, 61 to 90 and over 90 days late. Collection effectiveness compares what you collected with what you could have collected in a period. Add two working numbers: payments waiting to be matched, and disputes still open.

When you invoice on credit, debit accounts receivable and credit revenue, plus sales tax payable if you charge it. When the customer pays, debit cash and credit accounts receivable. A short payment leaves the difference open on the invoice until a person decides whether it is a dispute, a deduction the customer is owed, or a write-off. Your accountant sets the accounts each of those goes to.

It depends on your size and your accounting system, so we rank nothing. Small teams on QuickBooks Online or Xero can start with the reminders built into those systems. Teams that want invoicing and payments in one place look at BILL. Mid-sized and larger companies with an ERP look at Versapay, Billtrust, HighRadius, Upflow or Gaviti. We have not tested these products. The list on this page restates each maker's own page.

These are list prices from the makers' own pages, read on 22 Oct 2026. BILL runs from $49 to $89 per user per month, and its plans list automated payment reminders. Stripe Invoicing charges 0.4% or 0.5% per paid invoice. Upflow and Gaviti quote on request. Versapay, Billtrust, HighRadius, Paystand and Invoiced gave no price on the pages we read.

Check your accounting subscription before you pay for anything. QuickBooks Online sends up to three automatic reminders per invoice. Xero lists automated invoice reminders. Odoo has follow-up levels built in, though its letters and text messages need paid credits. NetSuite users can add Oracle's Dunning Letters SuiteApp. If those cover your reminders, the next thing worth paying for is usually the matching of payments, not more emails.

No. FactoryJet is a services company. We design, build and support an AI agent on your own accounting system, and you own what we build. There is no FactoryJet app to log in to. If you want a product to switch on this week, the eight listed on this page are a fair place to start, and we will tell you which one we would look at in your place.

Partly. Intuit says QuickBooks Online can send a first, second and third reminder, each up to 90 days before or after the due date, and only for invoices you have already emailed from QuickBooks. QuickBooks Online Advanced can use workflows instead. In the API, one payment can be applied to several invoices and credit memos, or left unapplied as a credit, which is where an agent does its matching.

Yes. Xero says its automated reminders nudge customers before and after the due date with your own message. For matching, Xero's AI feature JAX categorizes and matches bank transactions on the Growing plan and above. Xero's API applies payments to approved invoices, and a BatchPayments endpoint pays several invoices in one transaction. An agent works through that API for the payments JAX cannot place.

Yes, for reminders. Oracle's Dunning Letters SuiteApp sets up an automated dunning process and sends reminder letters by procedure and level. Oracle lists cases where a letter is not sent, such as paused dunning or no change in level. Oracle also says dunning procedures can be created only in the user interface, not by CSV, SOAP web services or SuiteScript. Our NetSuite AI agents page covers the wider picture.

With the Payment Reconciliation Journal. Microsoft says Apply Automatically matches bank statement lines to open entries using rules you set, first by name, bank account or address, then by document number, then by amount. Each suggestion gets a Match Confidence of Low, Medium or High, and a rule can require review. Business Central also sends reminders by level, with a grace period so a paid customer is not chased.

Yes. SAP's Service Layer, the API for SAP Business One, lists IncomingPayments, which it describes as payments from customers, and a DunningTermsService. Odoo's documentation describes follow-up levels by days overdue that can send an email, a letter or a text, and reconciliation models that pre-select matching entries. An agent reads and writes through those, and FactoryJet has worked on both SAP Business One and Odoo.

For bank transfers to Stripe invoices, yes. Stripe says it uses the transfer's reference code and amount to find matching invoices, and when several are eligible it takes them by due date. If it cannot reconcile a transfer, the money stays in the customer balance until someone does it by hand. Stripe can also email reminders for unpaid invoices. Payments outside Stripe need their own matching.

It can do the reading, matching and drafting. It should not make the decisions. In the agents we build, AI reads remittances, emails and short-payment notes, suggests which invoices a payment belongs to, and drafts reminder emails in your words. A person approves anything unclear, decides on credit holds and settles disputes. The built-in tools work the same way: Business Central marks low-confidence matches for review.

No. It changes where the hours go. Matching a payment with a clear invoice number was rule-based long before AI. What is new is reading messy remittances and customer emails. The work that stays human is the work that needs judgement and a relationship: deciding whether a key customer goes on hold, agreeing a payment plan, and settling a deduction your sales team half-promised.

For business customers, an agent can send reminders, answer "can you resend the invoice" emails, attach statements and flag accounts that need a call. In our builds it writes and sends email only, in the tone you approve, and it stops reminding as soon as a customer raises a dispute. Phone calls to customers stay with your team. Debts owed by consumers come under extra rules, covered in the next section.

Not one that can be trusted on its own. A chat app does not know your open invoices, your customers or your credit rules, and nothing checks its work. The same kind of AI model does useful AR work inside a proper setup: a limited sign-in to your books, written rules, a log of every email and match, and a person approving what is unclear.

The main federal one mostly does not. The Fair Debt Collection Practices Act defines a debt as an obligation of a consumer for personal, family or household purposes, so a business invoice to another business sits outside that definition. State laws and your contracts still apply, and selling to consumers changes the picture. This is a planning note, not legal advice. Ask your counsel before automating any collection from consumers.

It comes from Regulation F, the federal rule for consumer debt collectors. A collector is presumed to comply on call frequency if it calls a person about a debt no more than seven times in seven days, and not within seven days after a phone conversation about that debt. It is written for consumer debt. The agents we build for business invoices send email and leave calls to your team.

It depends on how many systems the agent reads and writes, and how many rules it follows. US agencies typically charge $5,000 to $15,000 for a single-workflow build, according to Layer3 Labs' 2026 guide, and $15,000 to $50,000 for several workflows. ProductCrafters puts custom AI agents of every kind at roughly $5,000 to more than $180,000. FactoryJet quotes a fixed price in writing after a short scoping call.

A pilot on one narrow job usually takes two to four weeks. For accounts receivable that is often cash application on one bank account, with every match waiting for a person. A production agent with reminders, disputes, permissions, logging and monitoring usually takes six to twelve weeks. Several entities, a second system for orders and lockbox files sit at the longer end.

Yes. Before a contract, we show working software on your own data. For accounts receivable that means a sample of payments and remittances your team has already applied, run through a first version of the agent. You compare its matches with what your team did, line by line. If your accounting system or a product would do the job as well, that is the moment we say so.

You own it. The code, the written rules and the test data are yours, and it can run in your own cloud account. FactoryJet quotes the build, and any support you choose, as fixed prices in writing. After launch we keep managing the servers, the AI models, the API connections and the upkeep, so your finance team does not have to. If you move the work elsewhere later, everything goes with you.