The barrier came down: AI readiness in wealth management
Week four of Beyond the Hype, a 12-week series from our CEO Charlie Symonds on what modern technology practically means for wealth managers and providers across financial services.
On 13 January 2018 the nine largest current account providers in the UK were required to do something no bank had ever volunteered to do. They opened their customers' account data, and the ability to initiate payments from it, to any authorised third party the customer chose, through a standard interface. The order came from the Competition and Markets Authority, following its 2016 investigation into retail banking, and the banks had spent most of the previous sixteen months building the plumbing. [1]
Almost nothing happened for a while. The first apps were clunky, consent journeys were awkward, and plenty of people inside the banks assumed that was the end of it.
By July 2024, open banking had 11.32 million users, and the CMA had confirmed the roadmap complete. [1] In June 2026, the CMA9 handled 40.16 million open banking payments; cumulative activity had passed one billion payments and 100 billion API calls. [2] The ecosystem was valued at over £4 billion in 2024. [1]
The firms that built many of the services customers now use on open banking data were not necessarily the firms required to provide the underlying interfaces. The barrier to building on bank data came down across the market. What determined who benefited was not access alone, but whether the organisation was arranged to use it.
Open banking matters because it shows what happens when a technological barrier falls: capability becomes widely accessible, but value accrues to organisations with the data, governance, skills and operating model to use it.
What has actually changed?
Over the past three weeks this series has looked at why the customer is hard to see, why the record is hard to trust, and why the regulator now wants proof from the data rather than from the file. This week is about the other half of the picture: what the technology makes possible now that it did not before, and why that has turned out to be the easier half.
A caution before going further. Nobody in this market needs telling that technology has become cheap, because it has not. Most firms I speak to have watched their IT spend rise every year for a decade, and the licence fees for the systems they already run are a standing grievance. What has changed is something different and more useful: the barrier to entry. Capability that a decade ago required a data team, a multi-year build and a vendor relationship only a life company could sustain can now be reached by a mid-sized advice firm or a provider's operations function, often with people they already employ.
The regulator said something close to this on 19 October. In a speech at Bloomberg, the FCA’s Chief Data, Information and Intelligence Officer, Jessica Rusu, described adoption as “not coming” but “here”. She cited a joint FCA and Bank of England survey, due to be published the following week, which found that 86% of surveyed financial services firms used AI in some form. [3]
The wealth market’s own figures tell a more specific story. The FCA’s 2026 wealth management survey, covering around 400 discretionary portfolio management firms, found that 13% used in-house or third-party AI tools, rising to 45% when firms considering adoption within twelve months were included. More than 92% outsourced at least one business function, with technology among the most outsourced areas. [4] The capability is therefore increasingly accessible, and many firms are already comfortable using third parties, but neither finding demonstrates readiness to change their operating model.
AI readiness in wealth management: where the gap is
Alirity has run an AI readiness assessment with organisations across sectors for a little over a year. Across six areas, the lowest scores sit consistently in people, training and capability, not in technology. Sixty-three per cent of respondents are not ready on people and training, with training scoring 2.06 out of 4. Fifty-five per cent are not ready on data. Technology, the area everyone expects to be the problem, scores higher than either. [5]
NextWealth's research on AI in advice, published in December 2025, lands in the same place from a different direction. Of 200 advice professionals surveyed, fewer than 5% worked in a firm with a full AI policy, guidelines and sign-off in place. Seventy-nine per cent cited compliance and systems integration as a barrier; 77% said the lack of a structured due diligence framework was one. The report's author summarised it in a sentence: many of the foundations simply are not in place. [6]
For a financial services leader, the decision on AI readiness in wealth management is therefore not simply which technology to buy. It is which customer or operational outcome to improve, what evidence will prove the improvement, and which changes to data ownership, controls, roles and governance are needed to make it repeatable. Technology selection comes after those decisions, not before them.
What it looks like from inside a firm
Consider an illustrative example: the operations director of a forty-adviser firm connects a modern data tool to a back-office export, a feed from the firm’s main platform and a spreadsheet of fees. Within a weekend, she has a reconciled view of every client’s holdings, fees and review status on one screen. The initial build is quick and inexpensive; embedding it is not.
Monday is harder. The paraplanners do not trust a number they did not key. The compliance manager wants to know who signed it off and against which policy, and there is no policy. Two of the advisers prefer their own spreadsheets. The managing director is pleased but asks who owns it when she is on holiday. By Friday the new view is still running, and the firm is still working the way it worked the week before.
None of those reactions is wrong. They are the organisation telling her, accurately, that it has not been designed to absorb what she built. The technology took nine hours. The operating model has not started.
Ten years apart
2016 | 2026 | |
Who could reach it | Life companies and the largest platforms | A mid-sized firm, a provider's operations team |
Skills needed | A data engineering team, a vendor integrator | Business analysts and operations people who understand the data, with a small amount of technical support |
Time to a first working view | Twelve to eighteen months | Days to weeks |
Dominant cost | Build and infrastructure | Licences, and people's time to change how they work |
Governance needed | A programme board | A policy, a named owner, and a decision about the system of record |
What stops it | Budget | Habits, policies that do not exist yet, and the question of who owns the outcome |
The comparison is directional rather than universal. In 2016, capabilities of this kind typically required larger specialist teams, longer delivery cycles and more bespoke infrastructure. In 2026, cloud services, packaged platforms and accessible AI tools can shorten the route to a first working use case. Integration, assurance, cyber security, resilience and data remediation can still be substantial.
The last row is the one to read twice. In 2016, budget and infrastructure were often the dominant constraints. In 2026, they still matter, but they are increasingly joined by harder questions of ownership, controls, adoption and operating-model change.
Why this matters to the business, not just the IT budget
The reasons to care are not abstract.
For an advice firm, a trusted joined-up record changes how many clients each adviser can look after well, which is the number that drives valuation when the consolidators come knocking. It reduces the professional indemnity exposure that comes from decisions made on stale data. And it changes whether good paraplanners and administrators stay, because the people most likely to leave are the ones spending their week re-keying.
For a provider, the same capability lowers the cost of servicing each policy, improves the adviser service scores that increasingly decide panel positions, and reduces the risk in the migration off the legacy book that has probably slipped at least once already. It also gives the board an answer to the fair value question in hours rather than weeks, which was the point of last week's article.
The fuller business case, including how to put numbers on all of this for a board, is the subject of week nine. For now the point is simpler: the value is real, and the thing standing between most firms and it is no longer the technology.
What I'm seeing
The paragraphs that follow are my own reading of the market, from the conversations I have with firms and providers, and should be weighed accordingly.
The most common failure I come across is not a bad tool. It is a good tool dropped into an operating model that was designed for a different era, with no decision about what changes around it. The pilot works, the demo impresses the board, and six months later the firm is doing exactly what it did before with one more subscription. Deploying new technology into old ways of working does not produce a new way of working. It produces an expensive old one.
The second is the assumption that readiness is a technology question. When a firm asks whether it is ready for AI, it usually means whether its systems can support it. The honest answer is almost always yes, and almost always beside the point. The questions that decide the outcome are whether anyone has written down what the firm's rules are, whether the data can be trusted, whether a named person owns the result, and whether the people whose work will change have been involved in designing the change. Those are organisational questions, and nobody has made them easier.
The third is that the human side of this is treated as a communications exercise rather than a design problem. The paraplanner who does not trust a number she did not key is making a professional judgement, not resisting progress. The adviser who keeps his own spreadsheet has a reason. A firm that wants the new capability to stick has to redesign the roles, the checks and the accountabilities around it, with those people in the room. That is slower than buying software, and it is the part that works.
Questions people are asking
What does democratisation of technology mean for wealth managers and advisers?
It means the capability to join up, trust and act on customer data, and to put automation and AI to work on it, no longer requires a scale of investment only the largest firms could carry. The tools, the hosting and the models are available to a mid-sized firm on terms it can reach, and the skills needed to use them are increasingly business skills rather than engineering ones. It does not mean the capability is free, or that it arrives configured for your firm.
Why do AI projects in financial services fail?
The available evidence points consistently towards organisational factors, not technology alone. Alirity’s readiness data identifies people, training and data as weak areas; NextWealth found fewer than 5% of surveyed advice professionals worked in firms with a full AI policy, guidelines and sign-off; and the FCA warned against retrofitting governance after deployment or simply automating existing processes. [3] [5] [6] Projects are more likely to fail when governance, trusted data, ownership and changed ways of working are missing.
Is legacy technology really the blocker?
Legacy technology is a genuine constraint, but it is rarely the only blocker. It shapes how data is captured, accessed, secured and reconciled, and it can materially affect cost and delivery risk. Yet a firm that cannot produce a joined-up view is often also carrying unresolved decisions about which system is authoritative for each fact. Replacing a platform without settling those ownership and data-governance questions can move the same problem onto newer technology.
What is an AI-ready operating model?
An AI-ready operating model has explicit rules that people and systems can use, trusted data with named owners, clear human accountability, governance before deployment, and roles redesigned with the people whose work will change. Technology enables that model; it does not substitute for it.
The client's side
A client never sees any of this. What they experience is whether their adviser arrives at the annual review already knowing what has changed or spends the first twenty minutes being brought up to date. Whether a question about a pension from an old employer is answered in the meeting or in a letter three weeks later. Whether the firm notices that their circumstances have changed before they have to say so.
Every one of those is an operating model outcome, not a technology one. The firm with the better tool and the unchanged ways of working delivers the same experience it always did.
Who's doing it well
The open banking story has one more lesson in it. The organisations that made the most of the opened data were those built, or rebuilt, around the assumption that the data would be available: small teams, decisions about data ownership taken early, and products designed from the customer's situation inward rather than from the account outward. That is an organisational description, not a technical one, and it is available to any firm willing to make the same decisions. [1]
Closer to home, the FCA’s description of good practice in its outcomes-monitoring review is worth rereading through this week’s lens. The strongest approaches began by defining the outcomes to be monitored, selecting data and indicators that revealed risks across the customer journey, and connecting governance to action and subsequent testing. Collecting metrics or producing management information was not enough: firms needed to explain what the evidence showed, what they did in response and whether the action improved outcomes. [7]
What this means for you
If you run an advice firm
The question to ask is not whether you can afford a modern data and AI capability, but whether your firm could absorb one on Monday morning. Write down who would own it, which policy would govern it, and which three roles would change. If you cannot answer those in an afternoon, that is the work, and it costs nothing to start.
If you are a product provider
The FCA data shows that outsourcing is already normal across the sector, with technology among the most commonly outsourced functions, but third-party access does not remove the internal constraint. Pick one servicing journey, decide which system is authoritative for each fact, define the outcome and controls, and redesign roles around the joined-up view before selecting or extending a platform. Vendors can provide capability; accountability for the operating model remains with the firm.
If you are a consolidator
The barrier coming down cuts both ways. The firms you acquire can now reach the same capability you can, and the ones that have organised themselves to use it will command a premium. Integration onto one platform is not the synergy. The shared operating model is.
In plain English: automation, RPA and agentic AI
Rules-based automation follows predefined steps. Robotic process automation applies that logic to screens and keystrokes, helping systems interact where direct integration is unavailable, but remaining sensitive to changes in interfaces and inputs.
Agentic AI differs because it can plan and execute a sequence of actions towards a defined goal using approved tools and data, within configured boundaries. It may adapt its next step to the context, but it does not remove the need for designed workflows, controls or human accountability. That combination of autonomy and uncertainty is why the FCA’s speech focused on accountability, escalation and “continuous, conditional assurance”. [3]
An agent acting on customer data needs explicit rules, trusted data, a named accountable human, monitoring and limits on what it can do without approval. Without those foundations, it does not create a faster firm; it accelerates uncontrolled decisions.
Discuss what this means for your firm
On Wednesday 18 November, 12:30 to 13:30, I will be taking that list apart with a colleague who has built and run agentic AI in production, in Taking Advantage of Agentic AI: what it is and is not, what it needs from your organisation, where the value sits, and where firms are getting hurt.
If you want a measured view of where your own firm stands across data, people, strategy and technology before then, the AI Readiness Assessment takes about fifteen minutes: Take the Alirity AI Readiness Assessment
If you would like to talk through what this means for your own firm book an informal conversation here: Book a chat with me
The open banking experience offers a useful warning for AI: opening the door creates opportunity, but it does not create organisational readiness. The firms that benefit will be those that decide how data, controls, ownership and work must change before scale arrives.
Open Banking Limited, “CMA confirms full completion of Open Banking Roadmap, unlocking a new era of financial innovation”, 9 September 2024. The article reports completion by the nine CMA-mandated banking providers, 11.32 million users in July 2024 and an ecosystem valued at over £4 billion. The CMA’s Retail Banking Market Investigation Order 2017 established the implementation framework; the PSD2 application date was 13 January 2018.
Open Banking Limited, “Open Banking surpasses one billion payments and 100 billion API calls”, 30 July 2026. The publication reports 40.16 million payments and 2.81 billion API calls in June 2026, with cumulative activity exceeding one billion payments and 100 billion API calls across the CMA9.
FCA, Jessica Rusu, "Building the future of finance in the age of AI", speech at the Bloomberg UK Policy Series, published 19 October 2026. "Adoption is not coming. It is here"; 86% of firms using AI per the forthcoming FCA and Bank of England survey; "AI should improve consumer outcomes, not simply automate existing processes"; "continuous, conditional assurance"; criticism of governance retrofitted after deployment. https://www.fca.org.uk/news/speeches/building-future-finance-age-ai
FCA, Wealth management survey report 2026, published 18 August 2026, data to 31 December 2024, around 400 firms. 13% using in-house or third-party AI tools, 45% including those considering within 12 months; more than 92% outsource part of the business, technology among the most common. https://www.fca.org.uk/data/wealth-management-survey-report-2026
Alirity, AI Readiness Assessment results, as published on alirity.com/overview. 63% not ready on people and training (training 2.06 out of 4); 55% not ready on data; 49% lacking strategic alignment; 52% with insufficient technology foundation. https://www.alirity.com/overview
NextWealth, "Clearing the path for AI in advice", December 2025, survey of 200 advice professionals, as reported by IFA Magazine, 2 December 2025. Fewer than 5% with a full AI policy; 79% citing compliance and systems integration as a barrier; 77% citing lack of a due diligence framework; "Many of the foundations simply aren't in place". https://ifamagazine.com/mind-the-ai-governance-and-due-diligence-gap-nextwealths-latest-report-highlights-learnings-for-adviser-firms/ and https://nextwealth.co.uk/research/clearing-the-path-for-ai-in-advice/
Financial Conduct Authority, “Outcomes monitoring: good practice and areas for improvement”, 27 July 2026. The review describes good practice across strategy and framework, data, management information and testing, and governance, oversight and culture. https://www.fca.org.uk/publications/good-and-poor-practice/outcomes-monitoring-good-practice-and-areas-improvement
