TL;DR
Get your next haul delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
Splitit CEO Nandan Sheth says AI agents should weigh a consumer’s broader finances—not just a monthly payment—when recommending whether to use pay later. Splitit said a test that included early-payoff details produced conversion about two to 2.5 times as high as a version with less information, but consumers’ willingness to let agents make credit decisions remains limited.
Splitit CEO Nandan Sheth says AI shopping agents need more than the lowest monthly payment to recommend whether a consumer should use pay later: they need context about cash on hand, existing obligations and total borrowing costs. In a PYMNTS interview, Sheth also said a Splitit test found higher conversion when shoppers received details about an early payoff option, a result that highlights how financial guidance could influence checkout decisions.
Sheth framed the issue around the agent’s responsibility to the shopper. “The agent is working for the consumer,” he told PYMNTS CEO Karen Webster. He expects early systems to present options for consumers to decide on, rather than independently accepting credit. “The starting point will be ‘surface me the offer, I’ll make the decision,’” he said, adding that context can help a person make a better choice.
The interview cited joint research by Splitit and PYMNTS Intelligence indicating that 61% of consumers would accept an AI recommendation about credit or pay later, while 2% would let the agent decide on its own. Those figures describe reported consumer attitudes, not observed usage or actual automated credit decisions.
Sheth described a Splitit test comparing two ways of presenting financing. One showed an $80 monthly payment, the APR and total cost. The other added that there was no prepayment penalty and explained that making lower payments for three months before paying off the balance could mean paying only part of the interest associated with a six- or 12-month term. Sheth said the second version produced a conversion rate about two to 2.5 times as high as the first. The interview did not provide the test’s sample size, dates or absolute conversion rates.
Why Payment Context Could Change Checkout
The proposal would broaden the role of shopping agents from finding a product or presenting a financing offer to comparing the offer with a person’s finances. An $80 installment may appear manageable on its own, but its suitability could change depending on a shopper’s available cash, other payment plans and the cost of paying over time.
That distinction matters because a recommendation can affect both consumer borrowing choices and merchant sales. Sheth’s test suggests that explaining repayment flexibility may influence conversion, though the reported result alone does not establish whether shoppers made better financial decisions or whether the effect would hold across products and customers.
For consumers, an agent that compares financing with immediate payment could help make costs and trade-offs more visible. But personalization would require access to sensitive financial information, creating a tension between useful advice and privacy. The interview presents this as a possible direction for the technology, not a description of a broadly deployed service.
As an affiliate, we earn on qualifying purchases.
From Product Search to Financing Advice
Sheth illustrated AI’s potential with a shopping example: while searching for a watch for his wife, an AI tool returned several versions of the requested product and suggested an unfamiliar boutique brand based on information it had about him and his family. He said he bought the alternative, which was roughly 10% to 15% cheaper. That example involved product selection; financing adds questions about repayment and affordability.
To answer those questions, Sheth said an agent might draw on information such as bank balances, credit-card statements, revolving balances or brokerage statements. He said an agent would not necessarily need every source, and that some financial data, supplied securely, could help it weigh pay later against a personal loan or immediate payment. He also acknowledged personal concern about breaches and misuse of data.
One privacy approach he raised was a personalized large language model that performs much of its processing locally and sends only specific tasks outside the consumer’s environment. He also suggested automated purchasing may be more acceptable for routine, lower-value spending. As an example, he described a grocery agent working within a $300 monthly budget and using a shopper’s preferences to find deals. These were examples and possibilities discussed in the interview, not confirmed product features.
“A dynamic relationship between agent and consumer”
— Karen Webster, PYMNTS CEO
buy now pay later installment plans
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The interview does not establish when or whether AI agents will be able to access the financial records Sheth described, how consumers would grant and revoke permission, or what safeguards would apply if data were exposed or misused. Sheth raised local processing as one possible privacy measure, but the report does not detail a specific implementation or independent security review.
It is also unclear how the reported conversion test was conducted. The article gives Sheth’s comparison of roughly two to 2.5 times the conversion rate, but does not state the participant count, test period, absolute rates or whether the result has been replicated. Conversion measures purchasing behavior; it does not by itself show that consumers saved money or selected suitable credit.
The cited survey figures likewise indicate stated willingness, not how people would act in a live transaction. The interview does not specify the survey’s sample, field dates or question wording. No timeline was given for when agents might make purchases or accept financing without a consumer’s final approval.
As an affiliate, we earn on qualifying purchases.
Consumer Consent Will Shape Adoption
The next steps will depend on whether companies can provide useful financial comparisons while giving consumers clear control over sensitive data and final decisions. Sheth’s stated expectation is that agents will initially surface an offer for a person to review, with greater autonomy more likely for routine purchases and defined budgets than for larger transactions. The interview offered no launch date or product announcement.
Further evidence would be needed to assess the test result, including its methodology, absolute conversion figures and whether fuller disclosures improve outcomes for consumers as well as merchants. For more autonomous payment decisions, questions include how an agent weighs borrowing costs against liquidity, how it handles conflicting information and when it must seek approval. The interview identifies these as developing issues rather than settled industry practices.
As an affiliate, we earn on qualifying purchases.
Key Questions
What does Splitit’s CEO say AI agents need before recommending pay later?
Sheth says agents should consider more than the monthly payment, including a consumer’s available cash, existing obligations and total financing cost. He discussed financial records as possible inputs but did not say every agent needs access to all of them.
What did the Splitit financing test find?
Sheth said the version that added information about paying off the balance early, with no prepayment penalty, had a conversion rate about two to 2.5 times as high as a version showing the monthly payment, APR and total cost. The report did not give absolute rates or the test methodology.
Are consumers ready to let AI agents choose credit for them?
The interview cited research saying 61% would take an AI recommendation about credit or pay later, while 2% would let an agent decide independently. These are reported survey responses, not evidence that consumers have handed agents that authority in practice.
What privacy concerns did Sheth raise?
Sheth said access to bank and investment information could improve recommendations, but acknowledged concerns about data breaches and misuse. He mentioned local processing as one possible way to limit how much sensitive information leaves a consumer’s environment.
When might an AI agent make a purchase itself?
Sheth suggested that consumers may be more open to automation for routine, lower-value purchases, such as groceries under a set budget. He gave no timeline, and the interview did not report a general service that autonomously accepts pay-later credit.
Source: rss
Evergreen bestsellers Picks
bestsellers
As an affiliate, we earn on qualifying purchases.
