Banks are increasingly cancelling artificial intelligence (AI) projects before they reach deployment, driven by heightened scrutiny over measurable returns on investment, according to new data from Infosys. Pre-deployment cancellations rose 33% between late 2024 and late 2025, even as 59% of deployed AI initiatives are delivering tangible business value, Dennis Gada, Executive Vice President and Global Head of Banking & Financial Services at Infosys, told The New Indian Express.
Rising Cancellations Reflect Stricter ROI Discipline
The surge in pre-deployment cancellations marks a shift in how financial institutions approach AI adoption. Gada said the increase reflected stricter scrutiny of projects before deployment, with banks placing greater emphasis on whether AI initiatives could demonstrate measurable business outcomes. According to the Infosys Bank Tech Index, initiatives cancelled before reaching production rose to 8,100 from 6,100 in the previous reading, while post-deployment cancellations fell from 3,700 to 3,400 over the same period.
This trend indicates that banks are not struggling more with AI implementation but are becoming far more decisive about which projects merit advancement to live environments. The data suggests a maturing approach where institutions are filtering out weak business cases earlier in the process rather than discovering failures after significant investment.
Three Key Problems Derailing AI Projects
Gada identified three primary reasons why AI projects fail to reach production: the absence of measurable unit economics, fragmented data, and treating AI as an IT project rather than changing underlying workflows. Projects lacking clear unit economics struggle to justify their costs against expected benefits, while fragmented data environments prevent AI systems from accessing the comprehensive information needed for effective operation.
Perhaps most critically, many banks continue to approach AI as a technology initiative rather than a business transformation effort. This IT-centric mindset fails to address the workflow changes necessary for AI to deliver value, resulting in pilots that never progress beyond experimentation. The pattern aligns with broader industry findings showing that unclear business cases and poorly defined ROI targets remain the primary barriers to AI scaling.
Deployed AI Shows Strong Performance
Despite the rise in cancellations, successfully deployed AI initiatives are demonstrating significant value. The Infosys Bank Tech Index reports that 59% of deployed AI projects are already delivering tangible business value, with customer service and cybersecurity showing the fastest returns. Software engineering applications are delivering the largest cost savings among deployed initiatives.
This performance gap between cancelled and deployed projects underscores the importance of rigorous upfront evaluation. Projects that make it to deployment typically start with narrow scope, have named owners, and are built on verified data foundations before business cases are signed off. In contrast, cancelled initiatives often stem from broad mandates to “do something with AI” without clear objectives or accountability structures.
Regional Variations in AI Scrutiny
The survey reveals notable geographic differences in how banks approach AI ROI. Europe shows the highest ROI scrutiny of any region, despite having the most mature data-privacy and governance framework in the sample, built on nearly a decade of GDPR compliance. This finding challenges the assumption that regulatory maturity creates bottlenecks for AI deployment.
Instead, European banks appear to be exercising greater discipline by questioning whether business cases hold before initiatives reach live environments. This represents a healthier approach than approving broadly and discovering failures later, according to industry observers. The pattern suggests that regulatory frameworks alone do not determine AI success; organizational discipline in evaluating business value matters more.
Broader Industry Context
The banking sector’s experience mirrors wider enterprise trends in AI adoption. A KPMG survey published in June 2026 found that 49% of large organisations had scaled back, narrowed, delayed or paused AI agent deployments after operating costs began to exceed value produced. Only 7% of surveyed senior leaders reported their organisations had reached established ROI from AI initiatives.
Gartner forecasts that more than 40% of agentic AI projects will be cancelled by the end of 2027 on cost and unclear value grounds. The research firm cites escalating costs, unclear business value, and inadequate risk controls as the primary drivers of these cancellations. In banking specifically, governance failures around audit trails, data quality, and unclear escalation rules are stalling otherwise sound projects.
Strategic Implications for Banks
The data suggests banks are entering a more demanding phase of AI adoption where boards increasingly scrutinise deployments for measurable gains in revenue, cost, and productivity. This shift is forcing institutions to move away from broad mandates to implement AI toward identifying specific business problems such as reducing delinquency rates, lowering cost per customer interaction, or improving first-call resolution.
For AI to transition from pilot to infrastructure, it must be deployed against measurable outcomes through systems that fit how enterprises actually operate. Banks that defined business goals before purchasing AI tools report better returns than those that did not, according to multiple studies. The pattern is consistent: financial controls and governance structures must precede deployment, not follow it.
Looking Ahead
Infosys describes banks as leaning into AI with more conviction than a year ago, but with real discipline—treating it as a core operating capability rather than a portfolio of pilots. This capability is meant to lift customer experience, engineering productivity, and platform modernization together. The 33% increase in pre-deployment cancellations alongside strong performance from deployed projects suggests this disciplined approach is taking hold across the sector.
As banks continue to refine their AI strategies, the focus on measurable unit economics, integrated data foundations, and workflow transformation will likely determine which initiatives succeed in production. The industry is moving beyond the experimentation era where ambiguity was tolerated, toward an accountability phase where only projects demonstrating clear business value advance.