Artificial intelligence (AI) in banking is often framed around ambitious new products or sweeping reinvention. Yet many banks have a more immediate use for it in supporting core processes. As an example, advisers still spend too much time on administration, client onboarding and searching for information held across different systems. A preliminary Finalix survey of more than 200 respondents across financial institutions in Europe and Asia points to onboarding and know your customer (KYC) data capture as leading sources of friction. For banks, that provides a practical starting point: find where time is being lost, rethink the workflow and apply AI where it will improve the work or strengthen control. Finalix was founded in Switzerland in 2001, opened its Singapore office in 2019 and now has more than 80 consultants. Andrew Rufener, Senior Manager, Data and AI, brings more than 25 years of technology and transformation experience, while Senior Manager Ronnie Chan has more than 30 years across private wealth management, investment banking and change delivery. They treat AI implementation as a business transformation, with data quality and ownership worked out alongside the technology. Employees still need the expertise to question the output and take responsibility for the result.
Key Takeaways
- Administration absorbs adviser time: Finalix’s preliminary survey suggests advisers are more burdened by onboarding, KYC capture and information preparation than by client interaction.
- Onboarding offers a workable first case: It combines visible client frustration, heavy manual effort, regulatory exposure and a measurable case for improvement.
- Process design comes before technology: Institutions need to work backwards from business, regulatory and user requirements before selecting a model.
- Poor data limits what AI can deliver: Faster processing does not make poorly governed or inconsistent records more reliable.
- Automation changes the work left to people: Automation can assemble evidence and draft reviews, but skilled employees must interrogate the output, resolve exceptions and remain accountable.
- Small senior teams can move across silos: Finalix favours compact teams that connect business and information technology (IT), carry work through implementation and transfer capability to the client.
- A small first use case should be built for reuse: A focused KYC use case can prove value without becoming an expensive programme, provided its data, governance and integration patterns can support further applications.
The Administrative Problem Inside Advice
Finalix structured its research around four stages of the investment-advisory process: understand, advise, execute and monitor. It asked where advisers spend the most time, why that effort arises and which activities they would be comfortable delegating to AI. Even at this preliminary stage, the answers put administration ahead of investment work as the main drain on adviser time.
“You might expect the biggest pain point to be portfolio management, investment selection or rebalancing,” Ronnie says. “Instead, the front office points to KYC data capture, onboarding and administration. That is very telling.”
Periodic KYC reviews, compliance requests, research and product information also appear in the workload. Every hour spent on those tasks is unavailable for client engagement, judgement or opportunity generation – the work for which relationship managers (RMs) are principally employed.
Why Onboarding Comes First
Onboarding sits at the intersection of revenue, client experience, operations, risk and compliance. For a complex high net worth (HNW) relationship, delays are particularly damaging because the institution may invest months of work before reaching a final decision.
“A client can go through nine to 12 months of process and still be declined at the end,” Andrew says. “That is painful for the client and for the bank.” The same process is also being asked to produce more structured, transparent and traceable evidence as regulatory expectations rise.
Finalix therefore uses KYC and client lifecycle management (CLM) as a practical test case for AI. In the model presented for the forthcoming webinar, document extraction, translation, data integration and rule-based checks perform much of the groundwork. Generative AI helps assemble the case and draft a readable report, with the compliance team reviewing, amending and approving the outcome. Finalix would integrate the solution with the bank’s existing environment, avoiding wholesale replacement of core systems.
AI as a Change Programme
Even a focused use case crosses organisational boundaries. Front-office, operations, compliance, risk and IT teams may own different parts of the same process, while external vendors add another layer. That puts the target operating model, decision rights and user adoption alongside the technical build.
“These AI projects are primarily change initiatives,” Andrew says. “The technology is new, but it is not the most complicated part. The complicated part is getting people to move and think differently.”
Finalix begins by mapping the current process and its pain points, defining what good should look like and identifying the gap. Business requirements are then translated into technical specifications and carried through testing, integration and implementation. Working backwards from regulatory, legal and user needs also prevents a technically impressive solution from solving the wrong problem.
Data Before Intelligence
Banks often reach the data problem first. Established institutions may have decades of records spread across platforms inherited through successive mergers and acquisitions. A smaller institution can sometimes move faster because its data is newer, its architecture is less fragmented, and its decision-making structure is lighter.
Ronnie warns against treating AI as a cleansing layer. “If the data is poor, AI does not fix it,” he says. “It accelerates it, polishes it and can make it look better than it really is.” A large balance sheet and familiar brand say little about whether the underlying records are ready. Institutions still need dependable sources, consistent definitions and a clear audit trail.
From Doing to Thinking
Automation also changes the skills a bank needs. If a machine collects documents, checks completeness and prepares a draft review, the employee’s job becomes reading the evidence, spotting inconsistencies, asking the right follow-up questions and taking responsibility for the decision.
“You take away a lot of the ‘do’ work,” Andrew says. “What remains is the ‘think’ work – and you need people who can actually do that.” Complex HNW cases may require more experienced reviewers after the routine work has been automated. Training, role design and knowledge transfer therefore belong inside the transformation from the outset.
A Smaller Delivery Model
Finalix uses compact project teams, bringing senior specialists together around a defined outcome. Their business, change and technical experience allow decisions to be made quickly. Fewer hand-offs also keep accountability with people close to the work.
“Our role is to bridge the gap between business and IT,” Ronnie says. “We work with technology teams and vendors to translate the business requirement into the technical specification and stay with it through testing, implementation, change management and integration.”
Andrew adds that the client should retain the capability to operate and develop the solution after the engagement ends.
Start Small, Design for Scale
KYC gives a bank a contained first problem with a visible owner, measurable effort and sufficient strategic value. The architecture should still be reusable. Data ingestion, governance, review and integration patterns developed for onboarding can later support periodic reviews, document generation, investment research or other parts of the advisory process.
“It is not simple, but it is doable, and it does not have to cost millions,” Andrew says. “The important thing is to start, learn and design for scale.” A working use case must reach the live workflow, improve output quality and be usable by the people accountable for the result. That is the difference between a project that reaches production and another demonstration.
The real test comes in production, when employees have to trust the output and reviewers have to trace the evidence behind it. A persuasive demonstration is only the beginning.
UPCOMING WEBINAR
AI in Banking: From Banking Front Pain Points to Practical Impact – Tuesday, 29 September 2026 | 3.00pm-4.00pm HKT/SGT | Online
Join Hubbis and Finalix for a one-hour webinar examining the findings from Finalix’s research and using know your customer and client lifecycle management as practical use cases. Andrew Rufener and Ronnie Chan will discuss where advisers lose time, which activities are suitable for AI support, why data and human oversight matter, and how institutions can move from isolated experiments to implementation. Further industry panellists will be announced.
Eligible attendees may earn one hour towards their continuing professional development/continuing professional training (CPD/CPT) upon completion of the short quiz and feedback survey issued after the session.
