A six-figure invoice is 90 days overdue. Automated reminders have been sent; emails remain unanswered, and the client now claims there is a dispute. At this point, another automated message is not going to help you get your money back.
AI now handles many reminders that businesses once sent manually. But reminders do not always lead to payment, especially in B2B collections, where an overdue invoice may involve a dispute, a payment plan, or an important client relationship. Collections become complicated when invoices are disputed, and customers stop answering calls or emails.
Recovering overdue accounts is not just about follow-ups. It requires real conversations, good judgment, and someone who knows how to move the situation forward without damaging the customer relationship.
So, can AI really replace the human judgment needed to recover complex or sensitive accounts? Or does effective collections still require a person on the other end?
Overdue Accounts Are a Structural Cash Flow Crisis, and AI Alone Is Not Solving It
Late B2B payments create cash flow challenges because businesses have to wait for the money they have already earned. When invoices remain unpaid for a long duration, outstanding accounts receivable limit the cash for payroll, suppliers, and operations. AI can help you send reminders, but billing errors, approval delays, and payment processes are not its job description.
Where overdue AR stands across industries in 2025 and 2026, and what is driving the increase
Overdue accounts receivable is no longer an occasional billing issue. PYMNTS 2025 data shows that 56% of small and mid-sized businesses are owed money at any given time, with an average of $17,500 outstanding per company.
Similarly, Atradius data shows that 40% of B2B invoices were overdue in North America in 2025, and 5% of long-overdue invoices were recorded as bad debts. In sectors such as US agri-food, 40% of B2B invoices were overdue and 5% were written off.
These numbers show that delayed payments are affecting businesses across industries and putting pressure on cash flow and working capital.
Why AI adoption in collections has not closed the recovery gap for complex and high-value accounts
Businesses are increasingly using AI for invoice reminders, account prioritization, payment follow-ups, and collection workflows. But using AI doesn’t ensure getting payments faster.
Simple overdue cases might be handled through automated reminders. But you need to have a different and reliable approach to manage complex accounts.
A complex account often involves billing disputes, missing information, approval delays, or multiple decision-makers that an automated reminder cannot resolve.
What the write-off rate on accounts past 90 days means for a business carrying average outstanding receivables
For a small medical practice or professional services firm operating on 10–15% margins, $17,500 in outstanding receivables is the gap between making payroll and missing it. It is a vendor invoice left unpaid while the firm waits for a client to respond to a fifth automated reminder.
When two or three accounts of that size age past 90 days simultaneously, the operational consequence is a working capital crisis that affects every function in the business.
AI Collections Tools Perform Well on Routine Accounts and Fail on the Accounts That Carry the Most Revenue
AI handles routine collections efficiently, but complex and high-value accounts often need human judgment. Specialists step in to negotiate payments and protect valuable client relationships.
What AI collections tools do well and where their documented capability ends
AI collections tools are genuinely good at what they are built for: automated dunning sequences, payment reminders, email, and predictive scoring of which accounts are likely to pay.
They can analyze historical data and let you know which customer will pay immediately, which needs a gentle nudge, and who can be a red flag. AI can also match incoming payments, invoices, and bank statements, which reduces manual booking work.
It’s a good option for debts under $1,000. Above that, the recovery rate starts to drop as accounts get more complex.
Large B2B customers need empathy and careful communication when something goes wrong with an invoice. A reminder sequence cannot resolve contract issues or genuine payment disputes.
A client relationship built over several years can end over a single tone-deaf reminder sent while a dispute is still unresolved. Issues like bankruptcy, court action, or other legal matters also require human judgment.
The account types and dollar thresholds where AI recovery rates drop and human specialists outperform
Higher-value invoices often need negotiation, and many eventually lead to litigation, which makes them quite different from routine accounts.
A complex account is not defined by dollar amount alone. There are other factors involved that make it complex, such as debtor disputes, financial hardships, contested payment terms, or a long-term client relationship that businesses don’t want to spoil.
B2B invoices above $25,000, or accounts having the aforementioned issues, require litigation or specialist human intervention rather than an automated touchpoint.
What happens to high-value disputed and relationship-sensitive accounts when only AI outreach is applied
Consider a $40,000 invoice owed by a long-standing B2B client flagged as disputed. An AI tool sends a standard reminder sequence, logs non-responses, and classifies the account as high write-off risk.
A trained human collections specialist calls the client directly, surfaces the dispute in conversation, and identifies that $15,000 of the balance is legitimately contested due to a delivery discrepancy. The remaining $25,000 is undisputed and payable.
The specialist negotiates a payment plan on the undisputed balance, documents the arrangement, and flags the contested portion for billing review. The full relationship is preserved.
The AI tool would have sent six more reminders and written off the account.
AI Collections Tools Create Systematic Compliance Risk That a Human Specialist Prevents
AI tools can scale collections, but at the same time they can also scale mistakes. Incorrect messages, contact errors, or poor timing can create compliance risks for the business.
What FDCPA and Regulation F require of any collections process and where AI tools create violation risk
The FDCPA and Regulation F limit how collectors contact debtors. They apply to both human and automated debt collection. For example, an AI system may create problems by sending too many messages, contacting someone at the wrong time, or exposing debt information to a third party.
It may also send messages even before required validation is completed. Automation does not remove these legal obligations.
How the same automation that scales outreach scales compliance violations
Automation tools can send thousands of messages at a time, and a small mistake in a message can spread as fast as messages reach customers’ inboxes. The CFPB’s annual FDCPA report shows the agency received 207,800 debt collection complaints in 2024, an 89% year-over-year increase. In 45% of the cases, consumers said they didn’t owe any debt at all. A pattern that CFB explicitly links to automated outreach operating at scale without human verification of account accuracy.
Regulation F also limits collection contact. Under the ‘7-in-7’ rule, a collector is presumed to be violating the law if it places more than seven calls about the same debt within seven days, or calls again within seven days of a conversation about that debt. This presumption applies specifically to phone calls and voicemails. AI can accidentally cross this limit when outreach is not tracked properly.
Before contacting an account again, a specialist can check for existing disputes and prevent a routine reminder from becoming a compliance issue.
The CFPB stated explicitly in 2026 that there is no “fancy new technology” carveout from existing debt collection regulations. AI systems are held to the same FDCPA compliance standards, UDAAP principles, and state-level consumer protection statutes as human collectors.
For a full picture of CFPB’s 2026 position on AI debt collection compliance, regulatory accountability sits with the business only. Most AI collections vendors disclaim responsibility for how their tools are used in their terms of service. The CFPB examines the outcome of the collections process, not the vendor’s configuration documentation, and civil liability under the FDCPA attaches to the creditor’s collections operation.
What state-level regulatory fragmentation means for businesses collecting across multiple jurisdictions
State-level consumer protection statutes layer on top of the federal rules, and they differ from state to state. The number of messages or calls allowed in one state can differ from what is allowed in another state.
Rules that are safe in one state can be a violation in another, which is exactly the kind of detail a specialist checks before the call goes out. Most automated workflows don’t handle these differences by default.
A Trained Human Collections Specialist Does What No AI Tool Currently Replicates in a Recovery Conversation
Negotiation, empathy, and escalation judgment are the specific capabilities that separate a recovered account from a written-off one.
What real-time negotiation in a collections context requires and why AI cannot replicate it
PwC’s Global AI Jobs Barometer found that new tasks in AI-exposed roles are 2.5 times more likely to require empathy, judgment, and creativity. It is exactly the capabilities that define effective collections on complex accounts.
C&R Software’s 2026 analysis of AI versus human agents in collections states explicitly that complex situations requiring negotiation flexibility make human collectors invaluable in ways AI cannot match. Such as with nuanced judgment and creativity, providing the flexibility needed to close challenging accounts.
It’s an ability to hear what a debtor avoids saying. A long pause on a call can mean many things. It could signal frustration, financial hardship, or a genuine dispute, and knowing the difference changes how the specialist responds. Automated responses can feel impersonal when a person is already frustrated.
How empathy and relationship preservation produce recovery outcomes that automated outreach does not
A collections specialist calls a client 90 days overdue. The client has lost a major contract and cannot pay the full balance immediately. The specialist identifies this in the first two minutes of conversation, not from a scoring flag, but from what the client says and how they say it. The specialist negotiates a structured payment plan on terms the client can realistically meet, documents the arrangement, and recovers 80% of the outstanding balance over 60 days.
The AI tool sent six reminder emails, logged six non-responses, and classified the account as high write-off risk. The difference is judgment about what the non-payment actually means.
What escalation judgment means in collections and why it determines which accounts get paid
Escalation judgment is a strategic decision-making process that is used to determine when an overdue account needs stronger action to get the payment.
Instead of treating every overdue account in the same way, businesses use escalation judgment to decide which accounts need to move from automated outreach to a specialist call or from a specialist call to legal referral.
This kind of collection escalation is what keeps a business from either giving up on an account too early or pushing too hard on one that just needs a different approach.
Factors such as repeated non-response, disputes, financial difficulty, and large balances help determine when a stronger action is needed. Contact rate and recovery rate are not the same thing; a tool that sends every scheduled reminder hasn’t collected anything if the balance is still unpaid.
Trained Human Collections Specialists Remain the Recovery Layer for the Accounts That Matter Most
The right staffing model for B2B collections is not AI vs. human. It’s AI for volume and trained specialists for the accounts where judgment decides the outcome.
What the correct collections staffing model looks like for a business with a mixed AR portfolio
An effective collections staffing model for a mixed Accounts Receivable (AR) portfolio relies on segmentation based on risk, value, and complexity. AI should manage high-volume, low-value accounts through automated dunning, reminders, and payment portals.
Trained human collections specialists should handle disputed, high-value, relationship-sensitive B2B accounts and customers facing financial hardships, where judgment, negotiation, and tailored follow-up can directly affect recovery.
Recovering a $30,000 overdue invoice from a client who then continues to place orders is a different outcome from recovering the same amount from a client who never does business with the firm again.
A human collections specialist who conducts that conversation with professionalism and empathy makes the former outcome possible. Automated dunning that escalates contact frequency without adapting to the debtor’s situation makes the latter more likely.
Why retaining trained human collections specialists is a cash flow decision, not a cost decision
Failed B2B payments cost businesses $118.5 billion per year globally. Rising interest rates, tariff disruption, and insolvency risk are driving more accounts into the complex, high-value category where AI tools produce the worst recovery outcomes.
Recovering just one additional $25,000 account in a month can cover a specialist’s cost many times over. The goal is not to replace AI, but to use it for routine accounts while human specialists focus on the balance where judgment can recover lost revenue.
How trained virtual collections specialists provide this capability without in-house hiring overhead
A trained virtual collections specialist can manage the recovery process end to end. They can review the AR aging report, prioritize accounts by value and complexity, document outcomes, and escalate accounts that need legal attention.
Businesses can add this support without the cost of hiring and training in-house staff. Trained virtual collections specialists available within 7 days provide this capability with no in-house overhead and are typically onboarded within one week of engagement.
Conclusion
AI can speed up collections, but automation alone cannot recover every overdue account. Simple reminders may work for routine balances, while disputed, high-value, or relationship-sensitive accounts require human judgment, negotiation, and careful follow-up.
A trained collections specialist can understand the reason behind non-payment, negotiate realistic solutions, protect client relationships, and decide when escalation is necessary. For businesses, the strongest approach is to use AI for routine outreach and human specialists for accounts where recovery requires judgment.
This combination helps businesses recover more revenue, protect cash flow, and avoid turning valuable customer relationships into unnecessary write-offs.
Most Frequently Asked Questions
What is the New York Health Information Privacy Act (NYHIPA)?
The New York Health Information Privacy Act (NYHIPA) is a state law designed to protect health-related data that falls outside HIPAA’s coverage. The original bill was passed by the NY legislature in January 2025 but vetoed by Governor Hochul in December 2025. Its revised version, S9269, was introduced in the 2026 legislative session and is under review. It will require businesses to get valid customer authorization before processing regulated health information.
Does NYHIPA replace HIPAA in New York?
No. NYHIPA does not replace HIPAA but covers what HIPAA does not. HIPAA is the federal law that governs protected health information (PHI) held by covered entities. NYHIPA covers a separate category of health-related data that is regulated health information (RHI).
Who does NYHIPA apply to?
NYHIPA can be applied to any business or organization that controls or processes regulated health information (RHI) belonging to a New York resident or someone in New York, even if the organization is based in other states. This can include wellness apps, wearable device companies, telehealth platforms, advertisers, employers, and other businesses that handle health-related information not covered by HIPAA.
What are the penalties for violating NYHIPA?
Under the 2026 revised bill (S9269), NYHIPA violations can result in civil penalties of up to $15,000 per violation, enforced by the New York Attorney General. The original bill included 20% of annual revenue, which was removed in the modified bill.
Is NYHIPA currently law in New York?
No. As of September 2026, NYHIPA has not been enacted. The original bill, S9269, was vetoed in 2025. The revised bill, S9269, passed both the Senate and Assembly but has not yet become law.
Does NYHIPA apply to remote healthcare workers?
Yes, if they handle regulated health information for a covered business. NYHIPA applies to organizations, but remote workers who handle RHI must follow the organization’s privacy and security requirements. Practices using virtual medical assistants must make sure their workers are trained on both HIPAA and NYHIPA requirements.
How does the 2026 revised NYHIPA bill differ from the 2025 version?
The 2026 revised NYHIPA bill expands exemptions, provides more specific definitions of regulated health information (RHI), replaces the earlier 20% revenue-based penalty with civil penalties of up to $15,000 per violation, and clarifies consumer rights and service-provider requirements. The revised bill also does not provide individuals with a private right of action.
