From Pilot to Production: Why Banking AI Projects Stall and How To Ensure Success

BIZTECH: What are the obstacles that lead to AI initiative failures in the finance sector? How much of the problem is technology, and how much is organizational readiness?

MORRISSEY: In financial services, the barriers are much more organizational than technological. The technology is maturing rapidly, but many institutions still face fundamental challenges with data, process and skills.

Common obstacles include fragmented or poorly governed data, integration challenges with legacy systems, difficulty measuring return on investment and a shortage of AI-specific skills. There is also often a disconnect between AI teams and business stakeholders, which makes it harder to embed solutions into real workflows.

The banks driving impact have the CEO as the chief AI officer. They lead from the top, and I’ve seen incredible collaboration inside the organizations. AI is helping bridge the gap between IT and business teams, fostering stronger collaboration and a more unified focus on delivering impact.

We recently released the 2026 Work Trend Index, which found that a growing share of workers across industries are using AI in advanced, resourceful ways. The problem is that most organizations aren’t keeping up. The constraint for most firms is the gap between what their employees can now do and what their organizations are built to support.

According to the Work Trend Index, organizational factors — things like culture, manager support, talent practices — account for more than two times the AI impact of individual factors like mindset and behavior, at 67% versus 32%. This means the defining question is not whether individuals have the skills but whether the organization has built the culture, management practices and talent systems that incentivize and support new ways of working.

READ MORE: Why does minimum viable data governance make sense for financial services?

BIZTECH: What are the biggest security and privacy concerns when moving AI from sandbox to production in a regulated banking environment?

MORRISSEY: The biggest concerns we consistently hear from financial institutions are around data privacy — ensuring customer information is protected and used in compliance with regulations; model transparency and explainability, particularly for decisions that directly affect customers, such as lending; and the risk of bias or unintended outcomes. There are also broader cybersecurity considerations, since AI systems can expand the attack surface and introduce new vulnerabilities.

This is where having a trusted, enterprise-grade platform becomes essential. With solutions and capabilities embedded across the Microsoft Cloud, institutions can apply consistent controls across identity, data and applications — helping ensure sensitive financial data remains protected while AI systems are operating at scale.

In addition, as organizations begin to leverage more advanced capabilities — such as Agent 365, which can autonomously orchestrate workflows and interact with systems — the need for strong governance becomes even more important. These systems sit closer to business processes and decision-making, so ensuring they operate within defined guardrails, from both a security and compliance standpoint, is critical.

What’s most important is that AI risk management is treated as a core part of the system design. That includes robust access controls, continuous monitoring, auditability and governance frameworks built in from the outset rather than added later. When those elements are in place, financial institutions can move from experimentation to production with confidence, knowing they can scale AI responsibly in even the most regulated environments.

BIZTECH: How can financial institutions increase the likelihood of AI success, including people, process and technology? Which should they prioritize?

MORRISSEY: The institutions seeing the most success are taking a balanced approach across people, process and technology.

On the people side, they are investing in upskilling and ensuring collaboration between domain experts and technical teams. From a process standpoint, they are redesigning workflows to incorporate AI rather than simply layering it on top of existing systems. And on the technology side, they are building strong data foundations and integrating AI into their broader platforms.

If there is one area to prioritize, it is aligning AI initiatives to specific, measurable business outcomes and embedding them deeply into workflows. Research shows that successful deployments tend to focus on narrow, high-impact use cases with clear integration into the business process.

BIZTECH: Are there certain AI use cases in banking that are easier than others to operationalize quickly? Which ones tend to generate early wins?

MORRISSEY: Many institutions are focusing on augmenting employees rather than replacing them — using AI to support decision-making or automate repetitive tasks. This approach helps build trust internally while demonstrating clear return on investment.

Here are a few examples:

• Banco Bradesco, one of the largest financial groups in Latin America, set out to streamline customer experience and internal BIA (Bradesco Intelligent Agent) focuses on digital customer resolution that is highly secure. Serving 74 million customers, 100% of initial interactions are handled by BIA, and it has achieved 83% resolution of customer inquiries.
• First Abu Dhabi Bank’s enterprisewide AI adoption is accelerating productivity, with Microsoft 365 Copilot deployed across the organization and supported by a library of more than 1,000 Copilot
• ABN AMRO Bank used Microsoft Copilot Studio to develop AI assistants for both customers and employees. The AI agent for customers now supports over 2 million text conversations and 5 million voice conversations every year, while the employee agent provides easier access to a wide range of IT-related and other internal resources.
• Wells Fargo is leveraging Copilot Studio to dramatically shrink the time it takes bankers to retrieve policies and procedures, moving from minutes to This helps bankers serve customers faster while improving consistency and operational efficiency across more than 4,000 branches.
• Westpac, one of Australia’s four largest banks, is using Copilot Studio to build several agents, including for HR and IT, to instantly answer common employee questions and handle simple tasks.

Click the banner below to sign up for HealthTech’s weekly newsletter.

BIZTECH: How do you see the financial industry adopting AI over the next few years?

MORRISSEY: We are entering a phase where the industry is moving from experimentation to scaled deployment. AI is no longer a stand-alone initiative; it is becoming embedded across core banking functions — from risk and fraud to customer engagement and operations.

Since the emergence of generative AI in late 2022, the initial wave largely focused on internal use cases. What we’re seeing now is the next phase, where agentic AI is moving beyond the enterprise and into direct customer engagement. For banks, that means shifting from inward-facing optimization to reimagining how they interact with and serve customers.

Banks will also need to prepare for the emergence of a new customer channel: AI agents. This includes today’s software-based consumer agents and, over time, more autonomous systems acting on behalf of individuals. Enabling this shift will require banks to modernize their platforms so these agents can engage securely, reliably and at scale.

Over the next one to two years, I expect the conversation to shift from “Where can we use AI?” to “How do we operationalize it effectively and responsibly at scale?” Institutions that succeed will be those that can integrate AI into their core systems with strong governance and clear business outcomes.

We’ll also see more advanced, agentic systems that can take action within defined guardrails. The key differentiator will not be access to AI technology, but the ability to execute — combining strong data foundations, governance and organizational alignment to deliver real business impact.