The adoption of artificial intelligence is galloping ahead across entire industries, promising all manner of process improvements and resource savings.
In the wealth management sector of financial services, there’s a conundrum. Namely, as evidenced in a 2026 KPMG survey, the biggest restraint on AI implementation is data readiness, with key information trapped and spread messily across multiple systems and documents. Yet the solution to this data fragmentation increasingly demands the application of AI to bring everything together into a single source of truth.
The challenge is to overcome this paradox. To vastly reduce the time taken for important tasks, such as financial reporting, from days of manual data handling to mere minutes through AI-led automation.
The state of play in the sector
The rate of AI adoption reported by financial services firms surveyed by the Financial Conduct Authority (FCA) and Bank of England is around 75%, significantly higher than the average across the UK economy.
The FCA expects AI deployment to continue to scale up. KPMG reports that around half of UK wealth and asset management firms are either experimenting with AI or in the process of establishing it. But, while 56% have a firm-wide AI strategy, only 11% have progressed to embedding AI capabilities in their core workflows.
KPMG also reports that those leading the way forward in embedding AI are scaling their investment to enhance their revenue generation, client engagement and client investment outcomes. AI investment is accelerating in the sector: 61% of firms report they allocate more than 3% of their IT/change budgets to AI, and 62% expect their AI budget to grow in the next 1-2 years.
The data bottleneck
With 83% of firms agreeing that their data readiness issues were the key lesson they have learned when striving to adopt AI, many organisations are evidently still stuck in this blockage.
In our conversations with customers in the wealth sector, we find that senior management are pushing their team to find AI use cases, partly so their firm is not left behind. AI budgets are often held by individual departments rather than allocated company-wide, which provides scope for running small pilots. The catch is that most firms’ data just isn’t ready for AI.
Operations directors tell us that standalone data management projects are hard to justify. They are expensive, can overrun, and deliver operational efficiencies rather than direct revenue. Data improvement work usually gets carried out on the back of something else, such as an investment platform migration, an acquisition or a transformation project. In each of these scenarios, poor data often turns out to be the bottleneck.
One firm we spoke to has around 70 systems in situ containing data, not all of them in use. With such fragmented technology, it’s untenable to simply point an AI tool at messy, conflicting sources of information, and expect it to know which ones are accurate or most valuable. Traditionally, data clean-up has been slow, manual and expensive, which is why it regularly gets put off.
Transforming data management via AI
With growing acknowledgement that AI adoption is hindered by legacy systems and
fragmented data, there’s positive news that firms are starting to point AI technology at the data problem itself. KPMG revealed interesting use cases among firms in this context, including dashboard and predictive analytics (61%), data quality and anomaly detection (56%) and data profiling (44%).
Investing in data foundations by harnessing AI means that data clean-up, which once called for awkwardly large and complex projects, can now happen faster and at lower cost.
An AI use case close to home
As a specialist data management firm operating in the wealth management arena, one of our challenges is to supply verified financial data to support UK tax reporting. Our provision of Excess Reportable Income (ERI) data helps wealth managers and financial advisers tackle one of the most tortuous issues affecting investors today: accurate tax reporting for offshore investments.
The need for this data is exemplified by figures from the lang cat Analyser showing that 93% of investment platforms offer offshore funds but only 55% provide ERI reporting tools for financial advisers and their clients. Platform support for ERI reporting has improved by just 3% since 2023.
This lack of support comes amid HMRC’s intensified focus on overseas income as it seeks to plug an estimated £59.2bn in underpaid tax, according to its latest assessment, imposing penalties for incorrect payment of tax as high as 200% of the tax owed.
Prior to adopting AI as part of our Managed Smart Data (MSD) platform, the extraction, ingestion and processing of this ERI data had been a manual process. It was heavily time consuming and dependent on a team with extremely high-level data handling expertise. Post-implementation of AI, we have reduced the turnaround time for reporting from one full working day to around 30 minutes. Humans still check for accuracy, in conjunction with AI, but this is a useful example of AI transforming data management in a concrete, measurable way.
What AI has brought to our table
With our MSD platform, data retrieval is AI-driven, automatically pulling data, documents and feeds from vendors and systems as soon as they become available. Our data extraction processes are also AI-driven, using advanced multi-agent technology to transform unstructured data into structured, business-ready data. The technology helps bring a firm’s tables, databases, spreadsheets and documents into one queryable place.
By incorporating what we call an ‘ontology layer’ in our platform, we add business context to clean data. AI agents can then interrogate and understand relationships between data, such as system-to-system and vendor-to-system, as well as between the data itself and the business. In lay terms, this functions like an instruction manual or data dictionary that explains the wealth management firm’s world to a machine. It covers who the clients are, what they buy, how products relate to companies, and what the data means and what it impacts.
Under the bonnet, this functionality automatically scans tables and writes plain-language descriptions of what each data column contains, such as dates, prices or market data. It builds a map of how entities connect: for example, company-to-products-to-clients-to-invoices.
In addition, to provide guardrails, it ranks information by how often it is used and referenced, as well as weighing who created a piece of information and how frequently it’s used. These vital guardrails help ensure accurate answers are supplied from patchy internal data, which matters in a sector where client reporting must be bang-on right.
AI-processed ERI data, made understandable by the ontology layer, enables a user to use common AI tools to ask in plain English which clients’ Capital Gains Tax reports need rerunning owing to tax liabilities for offshore income. This previously required using Structured Query Language to generate the sought-after data, a programming language that is often the reserve of developer and data engineers. As a result, the work involved drops from potentially weeks to minutes.
The bottom-line advantages are tangible and achievable. Significant efficiency gains have been generated for our platform’s users, with the data team redeployed to higher-value tasks such as client relationship management and product development, supporting a reduction in overall resource costs.
Resolving the data paradox with real-world consequences
By all accounts, it’s clear that the firms that pull ahead in their market won’t simply be the ones that adopt AI, they’ll be the organisations that apply it to their data foundations first.
In our highly regulated sector, the FCA stresses the importance of AI governance in its guidance to financial services businesses. In the regulator’s common-sense view, human accountability for AI-processed data matters as much as the accuracy of AI outputs. Across all industries, the clarion call to make sure people remain essential for judgement and responsibility, while AI efficiently tackles data extraction and analysis, will doubtless ring true.
