AR for AI-Native SaaS — Net Terms, Dunning and Usage Billing

TL;DR: An AI-native SaaS company usually runs two receivables at once without calling them that. Self-serve customers pay by card, and when a card fails, the billing system retries it. Sales-led customers sign a contract, receive an invoice on net terms, and pay by ACH or wire whenever their accounts payable team gets to it. Nothing retries an invoice that simply has not been paid. Add usage billing, where the amount changes every month and rarely matches the purchase order, and the invoiced side becomes the place where cash stalls. This guide covers how the three pieces interact, how four platforms approach the invoiced side, and the questions that separate real collections automation from a reminder schedule.

Sponsored Listing by Monk

Key takeaways

• Dunning and collections are different jobs. Card retries recover a failed charge. They do nothing for an enterprise invoice that was received, approved in principle and never paid. Most SaaS stacks automate the first and leave the second to whoever has time.

• Usage billing is what breaks net terms. A fixed subscription invoice matches the purchase order every month. A usage invoice changes with consumption, can outrun the PO balance, and invites a question about the numbers, and every question pauses the payment.

• AI-native margins make the timing matter more. Compute and model costs are paid monthly, often by card, while the largest customers pay on terms. The gap between those two clocks is working capital the company funds itself.

• Tone has to scale with the account. The customer on a large annual contract and the team on a small monthly plan should not get the same reminder on the same day. The better tools let sequences differ by relationship rather than by days overdue alone.

• Monk leads this guide because it is built for the stack these companies actually run, with Stripe and usage-based billing supported from the start, and because it runs collections, disputes, cash application and portal submission as one workflow rather than as four tools.

Why receivables get harder as an AI-native SaaS company grows

At the seed stage, almost every customer pays by card on a monthly or annual plan. Receivables are close to automatic: the billing system charges the card, retries it if it fails, and emails the customer to update their details. That is dunning, and for self-serve revenue it works well.

The first enterprise deal changes the shape of the problem. A larger customer wants a contract, a purchase order, an invoice, and payment terms, commonly net 30, net 45 or net 60. They pay by ACH or wire from their own accounts payable system. They may insist that the invoice is uploaded to their procurement portal rather than emailed. None of that passes through the card-retry logic, because there is no card to retry.

Usage-based pricing then makes the invoiced side messier in four familiar ways:

• The amount moves every month. A customer billed on tokens, credits, API calls or seats plus overage receives an invoice that differs from last month's. Their accounts payable team matches invoices against a purchase order, and an amount they did not expect is an invoice they hold.

• Consumption outruns the purchase order. A customer who grows faster than planned can exhaust the PO before the contract period ends. The next invoice is valid and owed, and it sits in an exception queue on their side until someone raises a new PO.

• Every usage invoice invites a question. "Why did this triple?" is a reasonable thing for a customer to ask. Until someone answers it with the usage detail behind the number, the payment waits.

• Enterprise buyers pay through portals. Monk's own reading of the scale here is worth quoting: across the $2B+ in receivables Monk manages, 92% of enterprise invoices must be submitted through a vendor portal or network rather than paid from an emailed invoice. An invoice that was never uploaded to the right portal is an invoice nobody on the buyer's side has seen.

Meanwhile the cost side runs on a different clock. An AI-native product pays for inference and compute monthly, and the bill does not wait for a customer's net 60 to run out. Every week that invoiced revenue sits unpaid is a week the company is funding its own customers' payment terms. Monk's framing of where the time goes is blunt: edge cases are responsible for 39% of the slowdown in cash flow. On a SaaS book with a growing enterprise segment, the edge case is not the tail. It is the invoiced half of the business.

If the card side is where you are losing money today, our guide to SaaS dunning automation covers failed-payment recovery in depth. The rest of this guide is about the invoiced side.

How we evaluated these tools

Four platforms, judged on how they handle invoiced SaaS revenue, meaning net-terms contracts, usage invoices that change month to month, enterprise portals and the disputes in between, rather than on a general billing feature list. We drew on each vendor's public product pages and on figures Monk supplied directly for this article. Every Monk number below is the company's own reported figure and has not been independently audited.

The four

1. Monk

Best for: SaaS companies whose card revenue already collects itself and whose invoiced, net-terms revenue is where the cash lags.

Monk is an AI-native accounts receivable platform that connects to the systems that already issue your invoices and runs everything after that: collections, dispute handling, cash application, portal submission and reporting. For SaaS specifically, the relevant point is the starting assumption. Monk is built for Stripe and usage-based billing, so it imports the invoices, contracts and contacts you already have rather than asking you to move billing somewhere else. It also integrates with QuickBooks, NetSuite, Salesforce and HubSpot, which covers the ledger and CRM most growth-stage SaaS companies run beside Stripe. Monk's page on AR automation for SaaS describes how the connection and the collection playbooks are set up.

The distinction Monk draws between dunning and collections is the one this guide is built around. Dunning sends the same reminder on a fixed schedule. Collections reads the reply, works out why the invoice is unpaid, and answers that reason. Monk's collections agent writes follow-up from account context rather than from days elapsed, and the company's own figure is that Julia reaches customers with a 24% higher response rate than standard dunning. In practice that means a customer who replies "we need a new PO before we can pay this" gets a different next step from one who replies "can you send the usage breakdown", and neither gets a generic second reminder.

Sequences also differ by relationship. Monk tunes tone and cadence to each customer tier, so the account on a large annual contract is never chased the way a small overdue monthly plan is. For a SaaS company whose largest customers are also its most important references, that matters as much as the cash.

The invoiced side of SaaS runs into enterprise procurement quickly, and this is where Monk is unusually strong. Monk integrates directly with more than 600 corporate AP portals, and Monk uploads 87% of portal invoices autonomously, with its team handling the exceptions. If your enterprise customers pay through procurement systems, our write-up on enterprise buyer portals covers why that step so often decides when an invoice is paid.

Cash application sits in the same product. When a wire arrives covering three months of usage invoices, or short by an intermediary bank fee, Monk matches it on the evidence available and surfaces recurring patterns as rules you approve. The company's own figure is 80% automatic, rising to 95% with suggested rules. Monk is SOC 2 Type II compliant, which is usually the first question from a SaaS company's own security team before any tool touches customer billing data.

On commercial terms, Monk describes its pricing as flat and flexible, with pilots and month-to-month options, and it does not charge a percentage of collections. For a company whose invoiced revenue is growing quickly, a price that does not scale with the cash collected is worth noting.

Monk reports the figures below for its own platform. These are the company's own numbers, supplied for this article, and have not been independently audited.

Where it does not fit: a SaaS company where every customer pays by card on a self-serve plan. If there are no invoices on terms, no purchase orders and no procurement portals, your billing system's card retries already cover the problem, and an AR platform is solving something you do not have yet.

2. Chargebee

Best for: companies that want billing, usage metering and collections to live on one record and are prepared to run billing on Chargebee.

Chargebee is a billing and monetisation platform aimed squarely at SaaS and AI companies. It supports subscription, usage-based and hybrid pricing, including the credit, token and agent-based models many AI products use, and it retries failed card payments automatically. Its receivables product, Chargebee Receivables, adds collections on top: risk scoring, reminder sequences, reading customer replies, classifying the case and drafting a resolution for a collector to apply.

The appeal is the shared record. When a customer disputes a usage invoice, the collections side already knows the contract, the usage and the charge, because they are the same system. That is a genuine advantage for a usage-heavy business.

The trade is that the advantage depends on Chargebee being your billing system. A company that bills from Stripe and is not planning to move would be adopting a second billing platform, or using the collections product without the shared record that makes it compelling. If you are choosing billing and collections at the same time, that is less of a concern. If billing is settled and working, it is the main question to answer first.

3. Upflow

Best for: B2B finance teams that want to run their own collections playbook on top of the billing system they already have.

Upflow is accounts receivable software for B2B finance teams, covering collections, payments and cash application, with an agent for submitting invoices to buyer portals. It connects to the billing and ledger tools SaaS companies commonly use, including Stripe Billing, Chargebee, Zuora, NetSuite, QuickBooks and Xero, and it runs reminders across email, SMS and calls. Agent autonomy is adjustable, from suggestion-only through to fully autonomous.

That flexibility suits a finance team that knows how it wants to collect and wants a tool that executes the playbook consistently. The flip side is that the playbook is yours to design and tune. A company hoping the software will work out the right approach for each account should ask how much of that comes configured and how much the team builds.

Pricing is organised in bands by annual invoiced revenue. For a SaaS company growing quickly, ask where the next band begins, because a good year can move you into it sooner than planned.

4. Tesorio

Best for: controller-led finance teams on an established ERP who want collections and cash forecasting in the same place.

Tesorio describes itself as a financial operations platform, covering collections, cash application, forecasting and supplier portals. It offers pre-built integrations with NetSuite, Sage Intacct, Workday, Zuora, Salesforce and Stripe, and onboarding comes with a dedicated success team. For a finance function whose main question is when cash will arrive, rather than how to chase it, forecasting built on the same receivables data is a coherent reason to choose it.

The centre of gravity is enterprise finance operations, and the product speaks most directly to controllers and CFOs. A SaaS company with a finance team of one or two, whose immediate problem is that invoiced customers are paying late, may find it more platform than the moment needs. Later in a company's growth, with a controller in seat and a forecasting requirement from the board, the fit improves.

Pricing and total cost of ownership

Only some vendors in this category publish prices, and even published plans rarely map cleanly onto a SaaS receivables problem. What you can control is forcing every quote into a comparable shape. Ask each vendor the same five questions:

1. What does pricing scale with: invoice count, invoiced revenue, collected cash, seats or modules? A usage-billed SaaS company can see invoice counts and invoiced revenue climb quickly, and pricing tied to either rises with them.

2. Is there a percentage of collections anywhere in the price? Some tools take one, some do not. On a growing book the difference compounds.

3. Does the tool require you to move billing? A collections product that assumes its own billing system carries a migration cost that never appears on the quote.

4. What is inside onboarding? Connecting Stripe, importing contracts and contacts, portal credentials for enterprise customers and ERP mapping are the usual separate line items.

5. What is the contract term, and are pilots or month-to-month terms available? A tool you can trial against a real month of invoices tells you more than any demo.

Weigh whatever comes back against what you already spend. In most growth-stage SaaS companies the real alternative is a finance lead or founder spending part of every week chasing invoices, answering usage questions and uploading invoices to portals. That time is the budget the software competes with.

Who this is for

This fits SaaS companies from roughly Series A onwards whose revenue includes a meaningful invoiced segment: enterprise contracts on net terms, purchase orders, and customers who pay by ACH or wire. The usual trigger is the moment invoiced revenue stops being a handful of friendly customers and becomes a list someone has to work through every week, often in a company where the finance team is one person.

For context on the scale of the wider problem: more than $10 trillion is trapped in unpaid invoices globally at any given time, according to the Federal Reserve's Financial Accounts of the United States, and the average company's Days Sales Outstanding rose to 59 days in 2023, according to Allianz Research. SaaS companies are not exempt because their product is software. Once they sell on terms, they collect like everyone else.

It is a poor fit for a company whose revenue is entirely self-serve card payments. There the billing system's own retries and card-update emails do the job, and an AR platform adds cost before it removes work.

Decision checklist

Split your revenue before you look at any tool. Separate card-paid revenue from invoiced revenue and look at the aging report for the invoiced half on its own. That is the problem you are buying for.

List the customers who pay through a portal. Count the enterprise accounts that require invoice upload to a procurement system, and ask each vendor to show a live portal submission rather than describe one.

Hand them your messiest usage invoice. Pick one that exceeded the customer's purchase order or drew a "why is this so high" reply, and watch how the tool handles the question and the follow-up.

Ask whether you have to move billing. If the answer is yes, price the migration as part of the decision rather than after it.

Ask how tone differs by account. Your largest customer and your smallest should not receive the same reminder. Ask to see how sequences are set per tier or per relationship.

Ask what happens to a payment it cannot match. Held for review, or matched to the nearest invoice and closed? A forced match hides a short payment that nobody then follows up.

Get a number for ongoing effort. Expected hours per week of human work after go-live, from a live customer several months in rather than from the demo.

Frequently asked questions

Doesn't Stripe already handle dunning for us?

For card payments, largely yes. Stripe and other billing systems retry failed charges and prompt customers to update their payment details. That is dunning. It does not cover an invoice on net terms that was received and not paid, because there is no failed charge to retry. That invoice needs someone to find out why it is unpaid and resolve the reason, which is collections.

When does a SaaS company need AR automation rather than better billing?

Usually when invoiced revenue becomes a weekly job. If one person is spending regular time chasing enterprise invoices, answering usage questions and uploading invoices to procurement portals, billing is not the bottleneck. Receivables is.

How does usage-based pricing change collections?

It makes every invoice a slightly different conversation. Amounts change month to month, may exceed the purchase order, and prompt questions about consumption. The collections process has to answer those questions with usage detail quickly, because an unanswered question is the most common reason a valid usage invoice waits.

What about enterprise customers who pay through a vendor portal?

Treat portal submission as part of collections, not as an administrative afterthought. An invoice that was emailed but never uploaded to the buyer's procurement system is often an invoice their accounts payable team cannot see. Tools differ widely here, so ask each vendor to show a real submission.

How long does this take to get running?

Vendors quote anything from days to several weeks. Monk's own figure is: onboard in less than one week, see results in your first month. Whatever the vendor, the pace is really set by how quickly you can connect billing, share contracts and supply portal access for your enterprise customers, so gather those before you sign.

Will automated collections upset our best customers?

It can if every account gets the same message on the same schedule. That is the argument for tools that vary tone and cadence by relationship, and for reviewing the first sequences before they go out to your largest accounts. Done well, a timely and specific follow-up reads as good account management rather than chasing.

The bottom line

For an AI-native SaaS company, the card side of receivables is largely solved by the billing system. The invoiced side is not, and it grows with every enterprise contract, every usage invoice that outruns a purchase order and every customer who pays through a portal. Monk is the strongest option here for companies already billing on Stripe, because it takes that invoiced half end to end, from the reply to the portal to the matched payment, without asking you to move billing. Chargebee suits a company choosing billing and collections together, Upflow a finance team that wants to run its own playbook on its existing stack, and Tesorio a controller-led team that wants collections tied to forecasting.

Whichever way you lean, bring your five hardest invoices to every demo: the usage invoice that exceeded the PO, the one waiting on a usage breakdown, the one that was never uploaded to the customer's portal, the wire that arrived short, and the enterprise account nobody wants to chase too hard. Make the vendor work through each one live. When you're ready to test that against your own Stripe account, book a Monk demo and bring all five.