MIT Technology Review describes banks’ use of agentic AI

MIT Technology Review describes banks’ use of agentic AI

MIT Technology Review published this sponsored Insights feature on September 4, 2025. It presents agentic AI as enabling large-scale automation of processes that rules-based approaches such as robotic process automation could not handle at that scale. EY’s Sameer Gupta says this could affect cost, efficiency, and customer experience. The feature names possible applications, including customer-service requests, loan approvals, bill payments, and extracting terms from financial agreements. It also cites a survey of banking executives. Respondents reported agentic AI use through deployments or pilots, and rated systems’ potential in areas including fraud detection and security. These are executives’ reports and assessments, not measurements of achieved results in the retrieved text. Citi’s Murli Buluswar argues that adopting new technical capabilities and changing how firms operate will distinguish firms that succeed from those left behind. He says people’s work will change meaningfully. The feature’s case for banking adoption rests on both potential process automation and the organizational changes its quoted leaders consider necessary. The survey offers a view of executive uptake and expectations, but the retrieved text does not establish what outcomes banks have achieved.

Publication: 7259ed22-b654-aeb3-00ef-c0a50d12cc6a

attribution — uncertain: The summary and limitations correctly identify this as sponsored Insights content rather than editorial work. However, the title, "MIT Technology Review describes banks' use of agentic AI", reads as the publication's editorial voice. The reading also never says the piece was produced "In association with EY", even though EY's own AI leader is a lead voice. Smallest repair: say "MIT Technology Review Insights" in the title and note the EY sponsorship.

This content was produced by Insights, the custom content arm of MIT Technology Review. It was not written by MIT Technology Review’s editorial staff.

summary — pass: The reading leads with the source's argument and names the actors (Gupta of EY, Buluswar of Citi). It explains robotic process automation as a rules-based approach, and agentic AI is defined beside the reading. It preserves Buluswar's survive-or-be-left-behind force. The question follows from the gap between pilots or assessments and measured outcomes. Softening "highly capable" to "potential" is acceptable given the caveat.

date — pass: September 4, 2025 matches the UTC day of originalPublished (2025-09-04T16:21:23Z). The page was edited on 2025-10-02, but the reading relies only on core article-body text. Later Deep Dive sidebar items, such as the AlphaGo ten-year piece, are neither used nor relied on. The source is a publisher page, not an archive copy, and no forecast is presented as an outcome.

support — pass: Each claim quotes the article body verbatim and keeps it attributed. The reading correctly frames the survey results as executives' reports, not measured results. It does not adopt the source's "already proving effective" framing as fact. Although the page was edited after publication, the cited passages are plainly the piece's main body. No unsupported benefits or costs are added.

More than half of executives say agentic AI systems are highly capable of improving fraud detection (56%) and security (51%).

media — pass: No media was selected, and no verified media candidates were supplied, so leaving media empty is appropriate.