Stuut Insights
AI in Accounts Receivable: What It Does and How It Works

Ben Winter
CBO
March 10, 2026

Table of contents
See Stuut in action
Get a personalized demo of Stuut and see how it can help with AR automation.
Get started
AI in accounts receivable (AR) takes on the follow-up work an AR team would otherwise do by hand. It monitors every open invoice, incoming payment, and customer reply. It sends reminders by email, text, and phone, answers customers' questions, matches each payment to the invoices it covers, and gathers the documents behind each dispute or deduction. That takes most of the routine work off the finance team and gives them time for the accounts and conversations that need judgment. Finance leaders decide how far the AI goes, and anything that affects finances, agreed terms, or customer relationships is handed off to a person on the team.
Why Accounts Receivable Work Is Hard to Automate
Much of the work in accounts receivable starts with something a person has to read, such as a customer's reply to a reminder, a remittance advice that lists what a payment covers, or a bank record showing that a payment arrived short. Each one leads to a decision, whether that is resending the invoice, splitting the payment across the right invoices, or choosing between chasing the customer and waiting.
Take a collections specialist on the AR team who opens a customer's reply and learns that the invoice is stuck because it went out without a purchase order number. A cash application specialist, the person who records incoming payments against invoices, runs into the same kind of problem when one payment arrives to cover six separate invoices and comes in short because the customer took a deduction on one of them. Before the cash can be posted, the specialist has to work out which invoice was short-paid and why. Neither task follows a fixed pattern, which is why traditional AR automation, which is built on rules such as sending a reminder 30 days after the due date, has never been able to take this work off a human accounts receivable team.
Stuut, the AI coworker for order-to-cash, changes that. It runs a team of specialized agents that read the same invoices, remittance emails, and portal messages your team reads. When a customer writes back that they never received an invoice, an agent resends it to the right contact and sets a new follow-up date, even though nobody wrote a rule or a template for that situation. The agents work through thousands of accounts at once, each with its own replies, payments, and open questions.
What AI Does in Each Part of Accounts Receivable
Accounts receivable has four main jobs: collections, cash application, disputes and deductions, and payments. The table below shows what the AI handles in each one and what your team still decides.
Part of ARWhat the AI doesWhat your team decidesCollectionsRanks open invoices by risk, contacts customers by email, text, and phone, and reads every replyWhich accounts stay with a person, and the tone and timing of outreachCash applicationReads remittances from emails, PDFs, portals, and bank files and matches each payment to its invoices, including partial and combined paymentsHow to handle the exceptions it flagsDisputes and deductionsSorts short payments and deductions by reason, gathers the documents behind each one, and routes the ones that need a decision to their ownerWhether to accept, write off, or contestPaymentsAdds a payment link to each reminder so the customer can pay straight from the messageWhich payment methods and terms to offer
Since these jobs are connected, AI works best when one system handles all of them. Say a customer replies that the payment went out on Friday. That one reply matters to collections, because the reminders to that customer may need to stop, and to cash application, because someone has to confirm that the money has arrived and which invoices it covers. A person would have to check both by hand. An AI agent that sees the full picture checks the bank file first, before that customer gets another reminder.
How a Team of AI Agents Works in Accounts Receivable
In a team of AI agents, each part of accounts receivable goes to its own specialized agent, and the agents pass work to one another on the same account. Stuut is built this way. Our AI coworker runs a team of specialized agents for more than 150 companies, including Fortune 50 and Fortune 500 enterprises. Four of those agents are live today:
- The Collections Agent decides which customers to contact and when, reaches them by email, text, and AI voice calls, submits invoices to accounts payable (AP) portals such as Ariba, Coupa, and Taulia, and reads every reply.
- The Cash Application Agent reads remittances, matches incoming payments to their invoices, including partial and combined payments, and posts eligible payments to SAP, Oracle, NetSuite, Microsoft Dynamics, and Sage. It flags the exceptions it cannot match for a person to review.
- The Disputes & Deductions Agent ties each deduction or dispute to its invoice, sorts it against the company's reason codes, and gathers the documents behind it so the owner can decide what to do.
- The Payments Agent adds a payment link to every email, text, and call follow-up so the customer can pay from the message itself.
All four agents work from the same account history, so every call, email, text, portal visit, promise to pay, and broken promise sits in one place. As a result, nobody on the team has to ask whether someone has already followed up.
How AI Automates Collections From Start to Finish
An AI collections agent stays with each invoice from the day it comes due until the payment is applied. In practice, the work runs through five steps:
- Prioritize. The agent ranks open invoices using payment history, aging, behavior, and risk signals, so the riskiest balances get attention first while the small invoices still get a reminder.
- Reach out. It contacts the customer on their preferred channel, in the tone your team has set, with the invoice and a payment link attached.
- Read the reply. It works out what each response means, whether that is a promise to pay, a request for a copy, a missing purchase order, a dispute, or an out-of-office message.
- Act on it. Depending on the reply, it resends documents, submits the invoice to the customer's AP portal, records a promise to pay along with its date, or routes a dispute to the person who owns it.
- Hand off or close. It watches for the payment, passes any broken promise to the person who owns the account, and stops as soon as the money is applied.
Most of the value comes from asking consistently and at the right time, and it takes less asking than teams tend to expect. In the collections data we published in What It Takes to Get Paid, half of collected past-due invoices closed within two asks, and nine in ten closed within six. What holds finance teams back is the time it takes to notice each stalled invoice and reach the person who can clear it, multiplied across thousands of invoices a month.
AI Collections Compared With Rules-Based Dunning
A rules-based dunning tool sends reminders on a fixed schedule and stops once an invoice is marked paid, which works well when customers have simply forgotten. The trouble starts when a customer writes back. Since a dunning tool cannot read replies, a customer who says they never received the invoice will keep getting reminders until someone on your team happens to see the email.
An AI collections platform reads those replies and acts on them, which is the work that fills most of a collector's day. Teams with low volume and straightforward customers can do well with dunning rules alone. Teams that deal with customer portals, deductions, several contacts per account, or a long tail of small invoices tend to outgrow them, because every reply the tool cannot read has to wait for a person on the team to pick it up.
How Finance Teams Stay in Control of AI
When finance leaders weigh AI in collections, the worry is what an automated message might do to a customer relationship. A badly timed reminder to a major account can cost far more than the invoice is worth. That is why control matters as much as automation. With Stuut, the finance team sets what runs on its own and what waits for a person's approval, every action is logged, and judgment calls go to a person.
In practice, our Collections Agent is built around those limits, in the product's own words: "You decide who Stuut contacts, how often, on which channels, and what needs your approval." Autonomy can be set separately for each channel and customer segment.
The largest accounts usually stay with people. In our data, only 5.4% of invoices over $250,000 ever received a collections reminder, compared with 16.8% of the smallest invoices, because relationships of that size tend to run through people who know each other. For those accounts, AI is most useful behind the scenes, as it can pull together the account history and open items before a human ever picks up the phone.
Escalations, disputes that need a decision, manual calls, and follow-ups on a broken promise to pay also go to a person. As Stuut's CEO, Tarek Alaruri, puts it in What It Takes to Get Paid, "The machine does the asking. People handle the fights, and only the fights." Every action the agents take is logged, so anyone on the team can open an account and see what was sent, when, and why.
How Much of the Work AI Handles Today
In our published data, 81.7% of outbound collection emails go out with no person involved. About a tenth are drafted by the Collections Agent and reviewed by a person before they are sent, and just 7.5% are written by a person from scratch. On the cash application side, the Cash Application Agent automates 95% of the work.
Automated asking does the most good at the slow end of collections, and that end is long. In our analysis, half of invoices were paid within 25 days and 70% by day 37. The next 20% took another 37 days, until day 74. The overdue money is also concentrated in a small number of invoices. In the same data, the largest tenth of past-due invoices at the typical company held 65.7% of the overdue dollars.
What to Ask About Data Security
Accounts receivable data is some of the most sensitive information a company holds. It includes customer contacts, payment terms, bank details, and the full history of who paid late and why. Before you give any AI tool access to your accounts receivable data, your finance team should have clear answers to these four questions:
- Is our data used to train AI models?
- Which security certifications and compliance standards does the platform meet?
- Who holds the encryption keys?
- Can we delete everything and prove that it was deleted?
Here is how we answer each of them at Stuut:
- Training. We do not use a customer's data to train models for other customers.
- Certifications. Stuut is SOC 2 Type II and ISO 27001 certified, and HIPAA and GDPR compliant.
- Encryption keys. We encrypt data at rest and in transit. Customers who add Stuut Shield, our paid security add-on, decide who holds the keys, either Stuut or their own team. As our security page describes, Shield tokenizes sensitive personal information "before it reaches the rest of the platform" and protects it with AES-256 encryption.
- Deletion. With Shield, revoking the key deletes the data. Every tokenized record becomes unreadable the moment the key is revoked, and the customer receives a deletion certificate for their compliance file.
How to Roll Out AI in Accounts Receivable
Integration matters most, because an AR tool that works outside the enterprise resource planning (ERP) system leaves the team maintaining a second set of books. Stuut connects to SAP, Oracle, NetSuite, Microsoft Dynamics, and Sage, among 27+ ERP and CRM integrations, so the team keeps one set of books.
Stuut also connects to Gmail, Outlook, Salesforce, and major phone and text providers. Setup is measured in weeks. Lift Solutions went live in 4 weeks, ZoomInfo in 5 weeks, and Bishop Lifting in 6 weeks across 45 branches.
Rollouts tend to follow the same sequence, and each finance team sets the pace. As our overview of AR automation describes it, "Most teams start with human review on everything, then open up autonomous outreach for lower-risk accounts over time." In practice, the rollout runs through six steps:
- Record a baseline. Note your days sales outstanding (DSO), the share of receivables more than 60 days past due, your cash application match rate, and the hours your team spends on follow-up and matching.
- Connect the ERP and inboxes. Ask the vendor to show you exactly which records the AI will see in each system and what, if anything, it will write back.
- Start with the work nobody has time for. The long tail of small invoices and the cash application exceptions that pile up at month end carry little relationship risk and are easy to measure.
- Review everything at first. Approve each message until the drafts are consistently right.
- Open up autonomy one segment at a time. Start with lower-risk accounts and keep the largest with people for as long as you want.
- Compare results after a full billing cycle. DSO will move last, because it still reflects invoices issued before the AI started working on them.
Example: How Lift Solutions Uses AI in Collections
Lift Solutions had an AR team of three people collecting across three disconnected systems, Sage, EVO, and TakeStock. After Stuut brought collections from all three systems into one place, Lift Solutions' average time to payment fell by 11 days, from 42 days to 31.
How to Build the Business Case for AI in Accounts Receivable
The case rests on cash that arrives sooner and on the hours the team gets back. Late cash carries a real cost. In our analysis, $1 million collected 2 months late is worth about $6,600 less than the same amount paid on time at a 4% annual cost of capital. To find the value a pilot needs to prove, apply that cost to your overdue balance and add the hours your team spends today on reminders and matching.
Measure both against the baseline you recorded before rollout. A pilot that shows faster payment on the small invoices it started with and fewer manual matches within the first billing cycle gives a finance leader the evidence to expand it.
Frequently Asked Questions
Can AI Do Accounts Receivable?
Yes, AI can now handle most of the routine work in accounts receivable, including prioritizing accounts, sending and answering follow-ups, matching payments to invoices, and sorting disputes and deductions. Most teams approve every message at first and let the AI send on its own once they trust it. Decisions that change money, terms, or a customer relationship stay with people.
Will Accounts Receivable Jobs Be Replaced by AI?
AI takes over the repetitive parts of accounts receivable, such as routine reminders and payment matching, while the judgment work stays with people. In our published data, escalations, disputes, manual calls, and broken-promise follow-ups are still 100% human-resolved. That leaves AR specialists more time for the accounts and conversations that need a person.
Can ChatGPT Do Accounts Receivable?
ChatGPT and other general chatbots can draft a collection email or summarize a remittance you paste into them. They do not work inside your ERP, bank files, or customer inboxes, and they cannot apply your approval rules or keep a log of every action they take. Accounts receivable depends on that work, which is why teams use AI built for AR.
How Do AI Agents Automate the Collections Outreach Process End to End?
AI agents automate collections outreach by ranking open invoices, contacting each customer on their preferred channel, and reading every reply. Each reply decides the next step, whether that means resending an invoice, recording a promise to pay, or routing a dispute. The agents keep following up until the payment is applied and hand disputes, escalations, and broken promises to a person.
Should I Use an AI-Powered Collections Platform or a Rules-Based Dunning Tool?
A rules-based dunning tool is enough for teams with low invoice volume and customers who mostly forget to pay. Teams that deal with customer portals, deductions, several contacts per account, or a long tail of small invoices are better served by an AI-powered collections platform. That kind of platform can read replies and act on them, which a dunning tool cannot do.
What Is Multi-Agent AI for Accounts Receivable?
Multi-agent AI for accounts receivable uses several specialized agents, typically for collections, cash application, disputes and deductions, and payments. The agents work from the same account history and pass work to one another. Stuut is the AI coworker for order-to-cash and runs a team of specialized agents that works this way.
What AI Tools Help Automate Accounts Receivable?
AI tools for accounts receivable range from rules-based dunning tools to AI collections platforms and systems that cover collections, cash application, disputes, and payments together. When comparing them, look at how deeply each one connects to your ERP, how much control your team keeps over what it sends, and how it protects customer data.
Is Our Data Used to Train AI Models?
Stuut does not use a customer's data to train models for other customers. Stuut's security page explains how data is encrypted, who holds the keys, and how it is deleted. Ask every AI vendor for the same answer in writing.

Ben Winter
CBO
Ben brings over a decade of go-to-market and operations expertise to building AR automation that actually works. He was VP Marketing at Fairmarkit (where he met Tarek) and GTM executive at Waldo before co-founding Stuut. He focuses on operations, product, and marketing, ensuring the platform integrates with existing ERP systems and delivers results in weeks rather than months.
Related articles

What is Days Sales Outstanding (DSO)? Definition, Formula, and Calculation
Tarek Alaruri
Mar 10, 2026

The Problem That Led to Stuut (And Why AR is Ready for Revolution)
Tarek Alaruri
Mar 10, 2026
See all articles
Setup time to learn more
About
Case Studies
Blog
Partnerships
Try AI Calling
Calculate Stuut’s Value
Terms of Services
Privacy Policy
© 2026 All Rights Reserved
Brand & Website by Day Job Web dev by Jomor Design
Based in NYC, backed by Insight Partners, Andreessen Horowitz, M12, and others