LC Ideas: Views & Insights
AI Rewards the Agile as Wealth Management's Old Advantages Erode
The competitive logic of wealth management is being rewritten. For decades, scale conferred advantage. Deeper research teams, broader distribution, and client relationships accumulated over generations. Artificial intelligence is dismantling that calculus.
"For the first time, it's a level playing field for organisations of varying scale," said Shilpi Chowdhary, Group CEO of Lighthouse Canton speaking on a recent panel discussion at the FT Live event in Singapore, titled "Which AI use cases are really delivering value for wealth management clients?"
"Earlier, you would have had tremendous resources with large organisations, but now you don't need that. Now you need agility, with which you can do much more."
Chowdhary was candid about where that pressure falls hardest.
Large institutions are carrying structural weight: legacy systems, accumulated risk aversion, and the institutional caution that comes with size.
"There is a lot of risk that a lot of large organisations are not willing to take. Organisations which adopt and adapt, at the same time, will make huge ground."
The client dimension of this shift is equally disruptive.
"Each one of the clients is now carrying an Einstein in his pocket. So how do you deal with that kind of information flow coming at you?" Chowdhary questioned.
Celine Le Cotonnec, Chief Data and Innovation Officer at Bank of Singapore, who was also on the panel alongside Chowdhary, approached the same disruption from the private bank side.
The barrier to entry for building technology has collapsed, she noted, a shift that now extends to the employees.
"You used to need a team of 20 to 30 developers in order to be able to develop an application. Now in a no-code environment, supported by AI coding assistant such as Winsurf or Anti-gravity, every employee can develop their own application by themselves, " said Le Cotonnec.
That creates a trust problem no institution can solve with technology alone.
"Who will the client trust the most? If we're not using AI in the way that we onboard the client, serve the client, or look at portfolio performance management, eventually the client will trust their own field AI. And that is not where we want to be, because this is a business of trust, and clients come to us for our expertise," she observed.
For both executives, the competitive moat of the future is not infrastructure or headcount but the speed and conviction with which institutions embed AI into the client relationship itself before the client builds their own but also equip Relation ship manager with efficient AI tools which are framed from the bank house view of the market and embedded our reflect on Strategic Asset Allocation.
Who owns the agent?
As agentic AI moves from pilot to production, the wealth management industry faces a governance question it has not yet formally answered: where in the organisation does accountability for AI agents actually sit?
"We're talking about the AI council. I don't see any HR representative in any of those councils, but there is a tremendous impact on skilling, population, and organisation design. What would we do with the agent tomorrow? Do we consider them as employees? Do they have a reporting in the org chart? Who is responsible for their monitoring, access control? Who is the maker, the checker of the agent? How do we ensure that agent organisation is not built in the traditional team silos but the agent organisation design is the most efficient in term of scale, access control given and specialised in their own task" said Le Cotonnec.
The question is not abstract.
It depends on the type of agent and how systematic its use is within the organisation, according to Le Cotonnec.
She also pointed to a structural divide that AI is now exposing.
"Traditionally, especially in banks, you have the business on one side and IT on the other, and there is a Chinese wall in between. The reality now is that we have moved into a much more self-service mode. People who are giving requirements are now able to produce code and eventually automate their own tasks, so the question is, how do we review that operating model and integrate it with the velocity required to adapt the market changes and client demands?"
Chowdhary argued that the governance question is fundamentally a leadership mandate, not a technology one.
“Any of this framework has to be top-down. To pass out the responsibility to an IT department and expect them to deliver, I don't think that could be a fair ask," he said.
His framing extended beyond governance into competitive positioning. As AI capabilities become commoditised and available to all firms regardless of size, the differentiator will not be access to the technology itself but the clarity of vision behind its deployment.
Building for what comes next
The speed of AI deployment has outpaced the security frameworks built to contain it.
"There are a lot of plugins, a lot of external tools that you can quickly add onto your existing systems — and there is no control on that," Chowdhary observed. Data, he noted, has become the primary target in this environment, given it is the oil of the future.
He placed responsibility for this squarely with chief executives, pushing back against the assumption that cybersecurity and AI risk sit within the technology function because if they don't, organisations will be in trouble.
He also pointed to the accelerating intelligence curve as a risk amplifier, noting that the trajectory from GPT-2 to GPT-4 suggests artificial general intelligence (AGI), AI systems with broadly human-level cognitive capacity, may arrive as early as 2027 to 2028, bringing a proportional expansion of exploitable vulnerabilities.
Le Cotonnec framed the risk from the inside, identifying workforce readiness as the more immediate threat. Management, she added, carries a clear obligation to address this at every level of the organisation, from relationship managers to senior leadership.
Both executives converged on a point the industry has been reluctant to state directly: the cost of deferring AI transformation on security grounds is no longer lower than the cost of moving forward with appropriate guardrails.
What is AI is delivering
Chowdhary offered a window into how Lighthouse Canton scales AI from experiment to production. The firm launched an enterprise version of Claude and tracked token usage across the organisation. The highest-engagement employees, regardless of function, were recruited into an internal AI lab and paired with engineers.
Once proof of concept was established, engineering teams were brought in to build the guardrails and production-grade infrastructure around it, a separation of roles he described as deliberate.
On the client-facing side, the gains are already tangible.
A company that previously sustained research coverage of 50 to 60 stocks can now extend that to 200, with output generated and vetted in a fraction of the time. Portfolio construction presentations can now be produced in a few minutes.
"We're looking at optimisation of portfolios, can I replicate that portfolio at half the cost? Can I change some of these funds with ETFs because they are low cost?" Chowdhary said, adding that the company is currently on version one of these tools and expects to reach version ten by year-end.
On the harder question of measurement, both executives acknowledged the gap between AI's observable impact and the industry's current ability to quantify it.
"Nobody is measuring it one to one right now," Chowdhary said. "But from idea to execution, that journey was very long . Now your prototyping can be so efficient and so quick, and that leads immediately into business production."
Le Cotonnec closed with a sharp challenge to how the industry currently frames AI value. "That's the pitfall; a lot of what we measure today is efficiency: you used to do that task in how long, and now you're doing it faster. We're not preparing for the future; we're not preparing for new types of skill sets"
She proposed a more demanding metric. "We always talk about the return on investment. We never talk about the cost of inaction, and that's something we're always missing. How much do you need to invest in your compliance for this efficiency to still be regulatory acceptable? Especially for client-facing use cases."
