Across life insurance, pensions and financial advice, organisations are investing heavily in AI capabilities to improve customer experiences, increase operational efficiency and unlock new growth opportunities. From personalised retirement guidance to AI-powered adviser support and intelligent customer servicing, the potential benefits are substantial.

However, while much of the industry conversation focuses on where AI can be applied, there is a more important strategic question that firms should be asking: where should AI not be applied?

The answer is critical. Although AI has a significant role to play across financial services, it is not relevant in every part of the operating model. In fact, some of the most successful future-ready organisations will be those that deliberately constrain AI from certain areas while enabling it to flourish elsewhere.

Life and pensions providers manage some of the most valuable and sensitive assets people own: their savings, retirement funds, investments and financial futures. These products often remain on administration platforms for extended periods, requiring absolute accuracy, resilience and regulatory compliance.

This is why the system of record must remain exactly that: a trusted, authoritative source of truth.

The system of record is responsible for maintaining customer policies, pension holdings, transactions, valuations and contractual information. It is the foundation upon which every customer interaction and regulatory process depends. When dealing with an individual's retirement savings or life insurance benefits, there can be no ambiguity, unpredictability or experimentation.

As a result, AI should not really be allowed, or treated with the utmost caution, to operate within the core system of record itself. These environments are simply too mission-critical. The consequences of inaccurate calculations, unintended decision-making or inconsistent outcomes are too significant when people's financial security is at stake.

Instead, firms should focus on implementing modern, next-generation and modular administration platforms that provide a stable and reliable foundation for the wider ecosystem.

A useful framework for understanding this approach is Gartner's Pace-Layered Application Strategy.

The strategy separates technology into distinct layers based on their purpose and rate of change:

Within a life and pensions operating model, the administration platform should firmly sit within the system of record layer. Its primary role is to accurately maintain and process policy and pension data.

Crucially, business processes should not be embedded deeply within the administration platform itself. Instead, process orchestration, workflow management and decisioning capabilities should sit within the middle layer, the systems of differentiation.

This architectural separation creates the flexibility required to innovate without constantly changing the core administration system.

When business processes are externalised from the system of record, organisations gain the ability to introduce technologies such as:

These technologies can interact with core administration platforms through APIs and services while leaving the underlying source of truth untouched.

For example, an AI-powered customer service agent may help a pension member understand their retirement options, summarise policy information or complete administrative tasks. However, it should retrieve and update information through governed processes rather than directly changing records within the administration system.

Similarly, financial advisers can leverage AI assistants to prepare for client meetings, analyse portfolios, generate suitability report drafts and identify client opportunities. The adviser benefits from enhanced productivity, while the underlying pension and policy data remains protected within the trusted record system.

This separation allows AI to do what it does best, augment decision-making, improve experiences and automate interactions, without introducing unnecessary risk into the core estate.

One of the biggest advantages of this architecture is the ability to accelerate innovation.

AI is evolving at an extraordinary pace. New large language models, AI agents, automation tools and customer engagement technologies are emerging almost monthly. Organisations that attempt to embed these capabilities directly into legacy record platforms often find themselves constrained by lengthy release cycles, complex testing requirements and significant implementation costs.

By contrast, when AI sits within the innovation and differentiation layers, new capabilities can be deployed rapidly without requiring changes to the administration platform.

This dramatically reduces time to market for new services and functionality. Firms can experiment, iterate and adopt emerging technologies without placing their core operations at risk.

More importantly, they avoid creating a future technology problem where today's AI becomes tomorrow's legacy platform.

For pension members, policyholders and advice clients, this approach delivers tangible benefits.

Customers gain access to increasingly sophisticated digital experiences, personalised communications and AI-powered guidance tools. Advisers gain intelligent assistants that reduce administration and allow more time to focus on client relationships. Operations teams benefit from greater automation and efficiency.

Behind the scenes, however, the core administration platform remains stable, trusted and highly resilient.

This balance between innovation and control is increasingly becoming a defining characteristic of market-leading financial services organisations.

Ultimately, the objective should not be to inject AI into every component of the technology estate. Instead, firms should build a future-ready operating model where every layer performs its intended role. This is aligned with The Pension Regulator's recent AI plan in relation to good governance, data quality and being able to foster responsible innovation.

The system of record should be modern, modular and best-in-class, acting as the trusted foundation for customer and policy data. Business processes should reside within a separate orchestration layer that enables agility and flexibility. AI, RPA, workflow technologies and future innovations should operate around this core, leveraging data and services while remaining decoupled from the underlying record system.

This architecture creates an operating model that can continuously adapt to technological change. As new AI capabilities emerge. whether generative, predictive or fully agentic, organisations can adopt them quickly without disrupting mission-critical operations.

The result is a permanently modern ecosystem: one that protects customers' savings and investments while continuously exploiting emerging technologies to enhance experiences, improve efficiency and drive growth. In the world of life, pensions and financial advice, that may ultimately be the most important use of AI, not replacing the foundation, but enabling everything around it to evolve.

Kevin Jones, UK Pensions Technical Lead – Festina Finance