AI chatbots in banking apps resolve blocked payments only 11.4% of the time

DECTA today released a consumer study on two back-to-back AI failures in UK banking and payment apps: the moment an app blocks a transaction, and the moment its chatbot fails to fix the resulting problem. The study draws on a 1,506-respondent UK consumer survey and 159,600 app store reviews across the 10 most-used banking and payment apps in Europe, including Revolut, Monzo, Wise, N26, and Starling Bank. Chatbots resolved the blocked-payment problem in only 11.4% of cases. Half the time (50.0%), users had to escalate to a human.

Key findings

• Blocked payments are near-universal: 77.2% of UK app users have had a payment, transfer, or transaction blocked by their app. 50.4% of those have had it happen more than once.
• Chatbots resolve roughly one in nine cases: 11.4% of users routed through a chatbot got the issue fixed. 50.0% had to escalate to a human, and 14.9% never got a resolution.
• Avoiding chatbots is difficult: 56.6% of affected users tried human chat first, but 76.3% ended up with a chatbot at some point during resolution.
• Trust in AI customer service is low: 65.2% of UK consumers trust a human agent most to fix banking app problems. 5.4% trust a chatbot. A 12-to-1 preference.
• Chatbot complaints are rising across Europe: Negative app store reviews blaming chatbots rose 55.49% year on year across the 10 apps analysed.
• Half of blocked transactions never complete: 48.0% of affected users said the transaction never went through, and 10.0% completed it on a different app or bank.

Why this matters now

In January 2026, the Treasury Select Committee warned that 75% of UK financial services firms now use AI and that regulation is not keeping pace. Earlier this year, Klarna reversed its decision to replace 700 customer service staff with AI after satisfaction scores fell 22%. In parallel, UK account closures hit a record 453,000 in the latest year, and the Payments Association has flagged aggressive AI fraud screening as a competitive risk for UK banks. DECTA’s data puts numbers on the consumer side of those headlines.

Methodology

The survey ran online from April 27 to May 4, 2026, among 1,506 UK adults aged 18+ who use at least one of the named banking, wallet, or payment apps and had used it in the past month. Age and location quotas matched the UK adult population. The app store analysis covered 159,600 reviews across Revolut, bunq, Monzo, Wise, N26, Sumeria, Starling Bank, Bourso, Monese, and Chase UK. The 34,167 one and two-star reviews were classified twice: first for chatbot-blame sentiment, then for the underlying problem. The 12 months to April 30, 2026 were compared against the prior 12 months.

What the data reveals

The problem is not chatbot adoption itself. It is the mismatch between where chatbots are deployed and what they can actually resolve. Among classified chatbot-blame reviews, four problem types account for 57% of complaints: identity verification, declined transactions, login or access issues, and account suspensions. These are decisions that need human judgement, account-level access, or real-time payment authorisation. The journey starts with an algorithmic block and ends with an algorithmic inability to explain it.

Additional findings

• 65.9% of UK consumers use two or more banking, wallet, or payment apps, averaging 3.35 apps per person.
• Block rates are nearly identical for single-app users (78.0%) and multi-app users (76.7%). Using more apps does not protect you from being blocked.
• Users who have experienced a bad block trust humans more (66.7%) than users who have not (58.4%). The block is what hardens the preference.
• Only 13.4% feel their app is over-cautious, despite 77.2% having been blocked. A 64-point gap between perception and experience.

Full findings: https://www.decta.com/company/media/ai-in-banking-apps-a-study-on-blocked-payments-and-chatbot-failure