Article | November 13, 2026
by Dhritiman Mukherjee, Managing Partner, Financial Services Industries, DXC Technology; Dr Marc Brogle, Principal Advisor, Modernization & Transformation, DXC Technology
Financial institutions are investing heavily in AI (projected to grow from $35 billion to $126 billion by 2028), yet most struggle to scale pilots and proofs of concept (POCs) into production. While banks and insurers successfully deploy low-value use cases such as chatbots, high-value applications like credit decisioning and perpetual KYC remain stuck in POC.
This AI value gap widens between leaders and laggards, driven by nine organizational and governance barriers rather than technology limitations. These include regulatory complexity, legacy system integration, talent gaps and risk compliance challenges unique to financial services. Closing the gap demands rebuilt governance structures, operating models and engineering practices specifically designed for enterprise-grade agentic AI deployment at scale.
So, the debate about whether AI works for simpler use cases in financial services is over.
Boards have reviewed pilots and POCs. Business cases are approved and budgets allocated. The main challenge facing banks and insurers now is scaling AI safely, consistently, profitably and in compliance with regulations across complex, interconnected organizations where reputation is critical.
In this first of two articles, we outline the scale of the opportunity, the growing gap between AI leaders and laggards and the nine key barriers to moving from pilot to enterprise-grade production. Our next article will examine each barrier and mitigation in detail, and explain how DXC supports financial institutions in closing the gap.
Financial institutions continue to increase AI investment. Global spending on AI in financial services is expected to rise from approximately $35 billion in 2023 to over $126 billion by 2028, a compound annual growth rate of 29%. Leaders at 84% of organizations report higher AI investment as confidence in the technology grows.
The potential benefits justify the investment. By 2030, AI could increase bank profitability by up to 30% and reduce costs by 30% to 40%. Early agentic AI deployments are already achieving 3% to 5% annual productivity gains, while scaled multi-agent systems can drive enterprise growth above 10%.
AI use cases span both front and back offices, including:
Most institutions can demonstrate successful AI pilots and POCs. However, far fewer have achieved enterprise-scale AI that’s integrated into core, high-value and complex processes, trusted by risk and compliance teams, customers and regulators, and delivering measurable business value.
Only about one in five financial institutions has deployed agentic AI into production, despite most actively using or evaluating it. Approximately two-thirds report difficulty in articulating a clear, viable business case for their AI initiatives. Over 40% cite implementation delays due to integration and scaling challenges. Industry analysts expect a significant share of agentic AI projects to be canceled by the end of 2027, not due to model performance, but because of rising costs, unclear business value and insufficient risk control frameworks.
Independent research consistently shows strong intent and real investment, yet a persistent gap remains between POC and production.
That gap between financial institutions’ aspirations and their current capabilities is widening. Few institutions are advancing from pilot to enterprise-wide deployment. As they do, their advantages increase: improved data, more reusable engineering assets, faster iteration cycles and governance that enables compliant deployment of higher-risk, higher-value use cases.
Everyone else is left running the same low-risk pilots on repeat. We call this the AI value gap: the growing divide between AI leaders who scale and realize compounding value, and those remaining in pilot mode. If unaddressed, this gap will increasingly distinguish industry leaders from others in financial services throughout this decade, reflecting the profitability advantages analysts already forecast for scaled AI leaders.
A key factor behind this gap is a common pattern in FSIs: institutions can easily deploy low-risk, low-value use cases, but often struggle to advance high-value, high-risk use cases from POC to production.
Simple use cases, such as drafting and templating, summarization, knowledge retrieval and general-purpose chatbots progress quickly due to low barriers. In contrast, complex use cases such as credit decisioning, dynamic portfolio rebalancing, KYC and onboarding at scale, post-trade reconciliation and continuous solvency monitoring, offer greater business value but face significant obstacles. As a result, most AI remains focused on low-value applications, while high-impact use cases remain in pilot.
This isn’t about model capability. It concerns implementing AI within highly regulated, complex organizations. Financial institutions face organizational, governance and structural barriers to scaling AI, not algorithmic ones. The same nine barriers consistently appear across banks and insurers, regardless of geography or size:
| Barrier | What Makes It Hard |
| Cost & ROI | Difficulty building the business case, proving ROI, and securing continued funding for AI beyond initial pilots. |
| Operating Model | No enterprise-wide model to productionize AI; ownership stays siloed in technology rather than the business. |
| Talent & Culture | Skills gaps and cultural friction between business, risk, and technology teams slow collaboration, adoption and scaling. |
| Risk & Regulatory Compliance | Evolving rules (EU AI Act, DORA, GDPR, model risk management) raise the bar for every deployment. |
| Data Fragmentation & Quality | Siloed legacy data, inconsistent schemas, and limited real-time access undermine model accuracy. |
| Legacy Systems Integration | Monolithic core banking and insurance platforms are difficult and costly to integrate with modern AI. |
| Trust & Explainability | Black-box outputs erode confidence among regulators, risk teams, and customers alike. |
| Security & Privacy | AI expands the attack surface; models, data feeds, and APIs all introduce new vulnerabilities. |
| Complex Ecosystem | A sprawl of vendors, platforms, and tools creates integration overhead, lock-in risk, and governance cost. |
While all industries scaling AI encounter these barriers, financial services face all nine at once, each intensified by the sector's unique characteristics.
Regulation is complex and continues to evolve. The EU AI Act, DORA, GDPR and established model risk management regimes such as SR 11-7 all apply. These regulations often overlap, span multi-jurisdictions and can be ambiguous when applied to autonomous agents. Legacy technology is also deeply embedded; core banking and policy administration platforms were built long before generative AI existed and cannot be easily replaced.
Governance obligations are exceptionally high. Model risk, operational durability and conduct requirements demand auditability and human monitoring, that most AI tools don’t provide by default. Operational complexity further increases these challenges. Large financial institutions manage multiple business lines, geographies and legacy systems simultaneously, each with its own risk appetite and control environment.
Put simply, financial services firms cannot choose between speed and security. They must achieve both simultaneously, in production and at scale.
Closing the AI value gap won’t come from better models or faster pilots. Financial institutions must rebuild key foundations: establish governance frameworks designed for agentic AI, not added retroactively; create operating models that integrate AI beyond IT; implement engineering practices and platforms that enable repeatable, production-grade deployment, and intentionally reskill and redesign the workforce for effective human-agent collaboration, rather than leaving adaptation to chance.
With over 40 years of experience in banking and insurance, DXC has managed mission-critical systems for many of the world's largest financial institutions. This expertise allows us to close the AI value gap by combining regulatory and risk knowledge, engineering capabilities and safe AI integration with legacy systems, supervisory frameworks designed for agentic AI and operating models that embed AI within the business for clear accountability.
We’ve supported financial institutions advancing high-value use cases such as trade reconciliation, treasury document intelligence, back-office operations and SME lending analytics from POC to production, delivering concrete results:
Enterprise value from AI is achieved by aligning the right technology with effective governance, engineering and operating models.
In our second article, we examine each of the nine barriers to scaling AI, providing additional evidence, insights into their causes and a functional framework for financial institutions to address them. We'll also present DXC's approach to closing the AI value gap: an AI scaling blueprint focused on governance readiness, use-case prioritization and repeatable, enterprise-grade engineering. This blueprint has helped clients progress from experimentation to achieving enterprise value.
Managing Partner, Financial Services Industries, DXC Technology
Principal Advisor, Modernization & Transformation, DXC Technology
BCG, Global Retail Banking Report 2025; Deloitte, State of AI in the Enterprise, Jan 2026; McKinsey, Agents for Growth: Turning AI Promise into Impact, Nov 2025; McKinsey 2025; Gartner 2025; NVIDIA, State of AI in Financial Services: 2026 Trends; Forrester 2025; G2 2025; Statista (ID 1446037); DXC AdvisoryX client data.
