Trustworthy AI in Financial Services

Governance and trusted data decide how fast financial institutions move AI from pilot to production. In 2027, five moves will separate the institutions that scale AI from those still stuck in pilots.

Artificial intelligence is moving rapidly up the financial services agenda. From improving fraud detection and accelerating regulatory analysis to transforming customer experiences, its potential is increasingly clear. The challenge for financial institutions is turning that potential into sustainable business value while maintaining customer trust, appropriate oversight and control. 

Over the past year, the conversation has moved beyond whether financial services firms should adopt AI. Financial firms now ask how to deploy AI responsibly, integrate it into operations and prepare for systems that act as well as advise. 

As guidance, in the UK developments from the Financial Conduct Authority (FCA), including its July 2026 Mills Review and its roadmap for Open Finance, highlight the scale of the opportunity and the importance of getting the foundations right. Across the Channel, the European Union have formed the AI Act’s high-risk obligations, including credit scoring, which apply from 2 December 2027. This is an addition to Regulation (EU) 2026/1744, in force since 27 July 2026. 

Across financial services, the details differ by business: a retail bank, an investment bank, an asset manager and an insurer face different AI economics and risks. However, five key metrics for success are common among them.

Five Priorities in 2027 for the Financial Services Sector

1. Establish a clear view of AI adoption. Understand where AI is being used, which applications are planned and who is accountable for their outcomes. Include both formally deployed solutions and experimentation taking place across business teams. 
2. Assess data readiness. Identify gaps in data quality, lineage, ownership, interoperability and permissions that could undermine AI use cases or limit the value of Open Finance. 
3. Match controls to risk. Ensure that testing, human oversight, access restrictions and ongoing monitoring reflect the potential consequences of each application, particularly where AI can influence customer outcomes or take action. 
4. Connect governance to business value. Prioritise use cases with clear objectives and measurable outcomes. Evaluate success through improvements in areas such as operational efficiency, decision quality, customer experience and risk management, alongside appropriate safeguards. 
5. Build for scale. Develop reusable data and AI capabilities, consistent standards and clear deployment processes so that successful use cases can be extended across the organisation without compromising control.

Turning Governance into an Enabler of AI Adoption 

Governance works best when built from the beginning and integrated into how systems are designed, tested and operated. In practice, this means establishing clear ownership of AI applications, understanding the data on which they rely, testing their performance against intended outcomes and monitoring them once deployed. Firms should also define when human intervention is required and how decisions or outputs can be challenged. 

The level of control should reflect the use case. An application that summarises internal documents presents different risks from one that influences lending decisions, handles sensitive customer information or initiates financial transactions. 

Effective governance also depends on the wider data environment. Inconsistent customer records, unclear data ownership and fragmented permissions can undermine the reliability of AI, regardless of the sophistication of the model. 

This is where data governance, architecture and engineering become essential to AI strategy. Reliable data foundations help organisations understand what information they hold, how it can be used and whether it is suitable for a particular application.

The UK Case: What the FCA is Signalling

As a leading global financial hub, the UK has to make decisions that lead by example and stay ahead of the curve. As such, the FCA has kept a principles-based approach to AI. Rather than introducing a separate, comprehensive rulebook for financial services, it relies on existing frameworks. Consumer Duty, the Senior Managers and Certification Regime, and current expectations around governance and controls remain central to how firms should manage AI-related risks.

The same logic applies to firms in any market: name an owner for every AI application, and be ready to explain what it decides.

Open Finance

In April 2026, the FCA published its Open Finance roadmap. Building on Open Banking, Open Finance aims to let consumers and businesses share financial data securely with authorised providers, potentially improving access to mortgages, lending, savings, investments and pensions. The roadmap prioritises collaboration and practical use cases during 2026, followed by framework design in 2027 and longer-term scaling between 2028 and 2030.

AI could help turn this broader access to financial information into more useful insights, but those benefits depend on the quality, consistency and appropriate use of the underlying data.

On 29 September 2026, the FCA published its research into Open Finance infrastructure. It stresses secure, trusted and scalable data sharing, with verification, delegation and accountability as AI services act for consumers and firms.

The Mills Review

Published in July 2026, the FCA’s Mills Review examined how AI could reshape retail financial services by 2030 and beyond. It identified four major areas of change:

• The transformation of firm operations 
• The evolution of consumer journeys 
• Changes to competition and market power 
• The amplification of fraud and cyber risks

The challenge grows as AI moves from assisting employees to carrying out multi-step tasks or acting autonomously within defined goals. Controls that suffice for an internal productivity tool may be inadequate for a system that influences financial decisions or initiates transactions.

Compliance sets the baseline. Controls proportionate to the risk let teams adopt AI with confidence.

Building the Foundations for Trustworthy AI

In 2027, the FCA moves from collaboration to framework design and the EU’s high-risk obligations take effect. The institutions with governance in place will shape both.

Talan supports organisations through the compliance journey, from data strategy and architecture through to engineering, governance, privacy and implementation. Our teams combine industry experts, transformation consultants and AI and data engineers; governance is built into how use cases are designed and run.

The firms that start now will be the first to reach production.