Defined Contribution (DC) pension assets are projected to hit £800bn by 2030.
On paper, that looks like a success story. In reality, 12.5 million working-age people are under-saving. Even those with substantial pots are struggling: 75% of over-45s lack a clear plan for accessing their funds.
Not all of that under-saving is an engagement problem. Some people, particularly those on lower incomes, simply cannot afford to contribute more. Automatic enrolment has helped, but minimum contribution levels remain too low for many. Where technology can make a difference is with the millions who are saving but are not equipped to make good decisions with what they have. The technology to close the "engagement gap" exists. The opportunity now is in applying it effectively.
With Pensions Dashboard connections going live in October 2026, the urgency to address the engagement gap is increasing. For the first time, members will be able to see all their pots in one place. That visibility will drive engagement and consolidation, and it will create winners and losers. Providers who use AI to personalise, educate and guide will be better placed to retain and attract assets. Those who don't risk losing them.
Most providers have spent the last three years refining the "digital baseline." We have the apps, the improved annual statements, and the sleek calculators. Yet, the industry is still seeing a familiar pattern: people log in, they look at the number, and they leave.
Why pension engagement is still low
Over the past few years, providers have done what you would expect. They have improved apps, simplified dashboards, modernised annual statements and introduced interactive calculators. In many cases, the digital experience is significantly better than it was even five years ago.
And yet the pattern is familiar. The members who do log in tend to check their balance and leave. Many never log in at all, and a significant number have lost track of pension pots entirely (something the Pensions Dashboard will help address).
Contributions rarely change, pot consolidation gets delayed, and when members do reach the point of accessing their pension, many do so without advice. Since the Pension Freedoms Act gave individuals full control over how they draw down their pots, the risk of poor choices has grown. Too many people withdraw too much, too early, and their savings don't last.
Members already have access to more information than ever. What they lack is the confidence to turn that information into a clear decision. Pensions are complex, long term and high stakes. The trade-offs are not always obvious, and once a choice is made it can feel overwhelming to change direction. When someone feels unsure, inertia becomes the safest option.
If engagement is going to improve, providers need to support decisions in context rather than simply present options. That is where AI, used properly, starts to matter.
The real blocker is data quality
Many pension platforms were never designed with AI reasoning in mind. Data structures vary across schemes. Historic decisions sit in PDFs or legacy systems. Lineage is not always clear. Core logic may be embedded in rules written years ago.
That works well enough for producing statements and running standard processes. It is far less reliable when an AI system is asked to interpret complex rules, explain tax implications or surface risks in real time.
If data is fragmented or inconsistent, AI will produce answers that sound plausible but lack context. If systems cannot share information cleanly, the model’s view of the member is incomplete.
Providers that are serious about AI are starting in a less visible place. They are investing in structured data layers, clear audit trails and consolidated records that AI systems can interrogate safely. The interface may look similar, but underneath it is supported by stronger foundations.
From conversation to capability
The industry conversation around AI in financial services has moved beyond chatbots. What's actually in production with customers largely hasn't. That gap is where the opportunity sits.
As the Pensions Dashboards Programme develops and Open Finance matures, providers will gain access to a more complete view of a member’s pension pots and wider financial position. Instead of relying on fragmented records held across multiple providers, they will be able to see how savings, income and liabilities connect.
With that broader picture, AI systems can begin to identify patterns across pots, contributions and income sources that would otherwise remain hidden.
Tone and personalisation matter, but they are not enough on their own. The difference lies in capability. Members need support that helps them complete complex tasks with confidence.
Designing for the guidance and advice boundary
Any discussion of AI in pensions has to acknowledge the regulatory context. The boundary between guidance and regulated advice is well established, and for good reason. The introduction of Targeted Support as a new regulated activity gives providers a clearer framework for offering personalised nudges without crossing into full advice. But providers still cannot allow AI systems to drift into recommendation territory without appropriate permissions and oversight.
Emerging approaches combine generative models with rule-based controls that validate and constrain outputs. Language that becomes directive can be intercepted. Scenarios that move beyond guidance can trigger escalation.
Building those controls into the design from the outset allows providers to offer richer, more contextual support while remaining within regulatory boundaries. It also reinforces an important principle: governance is not a layer applied at the end. It shapes how the system is built.
Rethinking the pension journey
Traditional pension journeys tend to follow a fixed path. Members move through predefined steps, often segmented by age or balance size. That structure provides clarity, but it does not always reflect lived experience.
Careers are less linear than they once were. Income fluctuates. Family circumstances change. External economic pressures influence financial decisions in real time.
An adaptive system can respond to those signals. Rather than relying solely on static thresholds or annual reviews, AI can help surface relevant considerations at the moment they become meaningful. The aim is not to overwhelm members with more data. It is to narrow focus to what matters now.
Trust as the foundation
Pensions operate on long time horizons. Trust builds slowly and is easily damaged. If AI feels opaque or overconfident, members will disengage. If it is transparent about assumptions, clear about limitations and supportive rather than directive, it can strengthen the relationship.
What this means for pension providers
The projected growth in DC assets marks a real shift in how retirement works in the UK. More money sits in defined contribution schemes, and more individuals are carrying the responsibility for complex financial decisions.
Digital upgrades have made pensions easier to see. They haven’t necessarily made them easier to navigate. The next stage is about embedding AI into the foundations of how pension systems work. That means better data, compliance built in from the start and tools that genuinely support people at the point of decision.
If you’d like to go deeper into the thinking behind this, download our whitepaper, Bridging the UK’s Pension Divide with Digital Solutions.
FAQs
Meet the author
Simon Hull is Head of Financial Services at CreateFuture. With over 20 years in banking and wealth, including UBS, Barclays, BlackRock and Deutsche Bank, he helps firms turn AI strategy into practical, accountable change.
Industry Insights
Explore the latest thinking from our industry and tech experts.
AI in UK wealth and pensions: 5 themes reshaping customer experience
How to scale with AI without over-hiring
Why do AI experiments fail and how do you scale AI-native delivery?