Artificial intelligence has moved rapidly from industry talking point to practical technology, but wealth management adoption remains uneven. Many firms are experimenting with proof-of-concepts, productivity tools and point solutions, yet few have rethought AI’s role across the full advice, servicing, onboarding and relationship management value chain. For Avishek Nandy, Partner at Bain & Company, the issue is not whether AI is capable enough. It is whether wealth managers have the leadership conviction, risk framework, cross-functional operating model and client-facing urgency to use it properly.
Key Takeaways
The bigger threat may come from clients, not competitors: Clients already have access to powerful AI tools and may use them to scrutinise products, pricing, fees and advice before wealth managers have equipped their own frontlines.
AI adoption remains shallow across wealth management: Many firms are using point solutions, but few are applying AI to end-to-end advice, servicing, onboarding or KYC workflows.
Enterprise usage lags technological capability: Nandy says the gap between what AI can do and how wealth firms are currently using it remains significant.
Path from use case to scaled transformation remains difficult: Large institutions often lack a clear strategic view of AI’s role across the enterprise and complexity in scaling the technology, while smaller wealth players face scale and investment constraints.
Risk can be managed systematically: AI introduces risks, but these can be mitigated through enterprise guardrails, observability, human oversight and existing execution controls.
Cross-functional silos are slowing progress: AI implementation requires business, technology and data teams to work together, particularly given wealth management’s reliance on unstructured data.
Adoption must be top-down: Nandy says AI needs CEO- or business-unit-led sponsorship because of the investment, risk, data, technology and operating model implications.
When the Client Knows More Than the Adviser
Not long ago, Nandy was sitting across from his own relationship manager, being pitched a set of sophisticated products. After the meeting, he put the proposal into an AI tool. Within minutes, the tool had broken down the products, assessed their historical track record, compared the fees against passive alternatives, and questioned whether the recommendations were suitable for his profile.
“The tool can break down the entire proposal,” he says. “The full cost picture – custody fees, transaction fees, management fees – laid out in a way that a client can interrogate before the conversation even happens.”
This is not a hypothetical. It is already happening, across client segments and geographies. Clients now have access to the same category of AI tools that wealth firms are still debating how to deploy internally. They are using those tools to scrutinise products, compare pricing, assess performance, identify cheaper alternatives and challenge the rationale behind advice — before, during and after meetings with their advisers.
For Nandy, this is the most urgent issue in wealth management AI — and the one least discussed. The debate tends to focus on internal productivity, cost efficiency and workflow automation. Those matter. But the more immediate pressure point is the client. The information asymmetry that has underpinned the adviser-client relationship for decades is eroding. Wealth managers that have not equipped their frontline accordingly are already exposed, whether they know it yet or not.
“You are leaving your frontline exposed,” he says. “The bar is getting higher in terms of what customers will expect.”
The Gap Between Talk and Transformation
Nandy sees no shortage of AI discussion across wealth management, but far less evidence of meaningful transformation. Many firms can point to pilots, internal productivity tools or narrow use cases, such as summarising CIO reports or supporting relationship managers with discrete content tasks. For him, those are largely hygiene factors.
The more important question is whether institutions are using AI to rethink the full advisory and servicing model. That includes advice generation, client engagement, onboarding, KYC, product explanation, pricing transparency, portfolio construction and frontline enablement. On that measure, he sees the industry as still early.
“We are still in the early stages of meaningful enterprise-wide transformation,” he says. “There is progress, but the gap between what the technology enables and how it is actually being deployed remains significant.”
That gap matters because enterprise AI has already moved beyond generic chatbot experimentation. With the right architecture, controls and guardrails, Nandy sees AI as capable of supporting complex advisory and workflow problems in a much more structured way than open-ended consumer tools.
For wealth management, the potential is particularly material. The industry has spent years trying to solve issues around advice quality, relationship manager productivity, product relevance, compliance, personalisation and scale. Properly deployed, AI can help address many of those long-standing constraints.
Why Institutions Are Holding Back
Nandy identifies several reasons why adoption has been slower than the technology itself would suggest. The first is strategic clarity. For large universal banks, AI cannot be considered only within the wealth business. It has to sit within a broader enterprise view of how the group will use AI across functions, businesses and risk domains.
That creates a leadership dependency. Without a clear board and management position on AI’s role across the organisation, wealth teams may pursue bottom-up initiatives, but struggle to scale them into business-changing platforms.
“There is a lot of bottom-up activity happening,” he says. “Wealth teams may be helping RMs create small tools, but it is not at the level of scale that can make a massive difference.”
For more focused wealth managers, the challenge is different. The issue is not necessarily organisational complexity, but scale. Building AI properly requires investment in technology, data, architecture, risk controls, operating processes and adoption. For smaller firms, the question is whether they have the scale to justify and sustain that investment.
The result is a market with many exploratory initiatives, but fewer enterprise-grade deployments. Business teams are often interested and usually understand the problem statements they want to solve, but the path from use case to scaled transformation remains difficult.
Risk, Guardrails and the False Leap of Faith
Risk appetite is another major constraint. Institutions worry about hallucination, unsuitable recommendations, uncontrolled outputs, regulatory exposure and the optics of technology appearing to replace people.
Nandy does not dismiss those concerns. Instead, he argues that many firms stop too early in the risk conversation. They identify the risk, but underestimate the extent to which enterprise AI systems can be designed to manage it.
“This is not about asking LLMs to provide advice,” he says. “You build enterprise systems with the right guardrails.”
Those guardrails can include evaluation frameworks, observability, controlled data access, model monitoring, workflow permissions, human-in-the-loop review and escalation processes. Execution can also continue to pass through the same checks and balances that govern existing investment processes.
For Nandy, the key is to approach AI risk systematically rather than emotionally. The risks are real, but they are identifiable and manageable. Wealth managers already operate in a domain where every action carries some form of financial, regulatory, conduct or suitability risk. AI should be treated through the same discipline.
“People get stuck at, ‘it is risky’,” he says. “But there is a way to manage those risks. It is not a leap of faith. There is quite a lot of science in how you can manage it.”
The Cross-Functional Challenge
Even where the business case is clear, AI implementation is structurally difficult. It cannot sit solely with the business, technology, data science or operations. It requires all of them to work together.
Nandy says business teams are often the most enthusiastic because they understand the commercial and client-facing problems. They know where relationship managers lose time, where onboarding breaks down, where product explanation is inconsistent, where advice quality varies, and where clients experience friction. But the business cannot solve those issues alone.
AI depends on data infrastructure, technology architecture, security, model design, governance and integration into workflows. In wealth management, that is especially challenging because much of the relevant information is unstructured. Client conversations, post-call notes, adviser observations, investment preferences, relationship context and behavioural signals are not always captured cleanly in traditional systems.
“AI is just a layer,” he says. “You still need the data underneath it.”
Historically, banks have become more comfortable managing structured data. Wealth management, however, is a knowledge business with a large volume of unstructured information. Relationship managers may know their clients deeply, but much of that knowledge remains in notes, conversations or individual memory. Turning that into usable enterprise intelligence requires data, business and technology teams to work in an integrated way.
Where those teams remain siloed, AI progress slows. Firms may build promising tools, but struggle to embed them into daily workflows, connect them to the right data, or make them usable at scale.
The Need for Top-Down Leadership
Because of this complexity, Nandy believes AI adoption has to be led from the top. It cannot remain a collection of disconnected pilots or enthusiasm from individual teams. The investment requirements, organisational changes, risk considerations and cross-functional dependencies are too significant.
For wealth managers within larger groups, that leadership may need to come from the CEO or executive committee. AI must be treated as a strategic operating model issue, not a technology side project.
“It has to be CEO down,” he says. “Because of the cross-functional nature, the investment required, the risk management required and the complexity of bringing different people together, it has to be led by the top.”
Top-down leadership also matters because AI adoption forces prioritisation. Institutions need to decide where AI will create the most value, which workflows should be redesigned first, how to govern usage, how to train the frontline, and how to sequence investment. Without senior sponsorship, firms risk spreading effort across too many small use cases without creating material impact.
The Client-Side Risk
Nandy does not see wealth managers facing immediate competitive consequences simply because peers are deploying AI faster. The industry is still in a hesitant phase, with most institutions watching each other closely. A few players are moving ahead, but many remain cautious.
The larger risk, in his view, is not what the bank next door is doing. It is what clients are already doing.
Clients now have access to powerful AI tools that can test, question and deconstruct advice. They can upload product material, compare pricing, analyse historical performance, assess fees, identify cheaper alternatives and challenge the rationale for recommended investments. That shift changes the standard expected of relationship managers.
For wealth managers, the implication is clear. If clients are using AI to scrutinise advice, the frontline cannot be left without equivalent or better institutional tools. The best relationship managers may already be adapting on their own, but relying on individual initiative creates inconsistency and leaves the broader adviser base exposed.
“You are leaving your frontline exposed,” he says. “The bar is getting higher in terms of what customers will expect.”
Rethinking the AI Question
For Nandy, the central issue is not whether AI will replace advisers. It is whether wealth managers will equip advisers to remain credible in a more transparent, more analytical and more demanding client environment.
The firms that make progress will not simply add AI tools around the edges. They will redesign workflows, connect data, build enterprise guardrails, align leadership, train frontline teams and embed AI into the operating model of advice and servicing.
That requires more than experimentation. It requires discipline, investment and ownership.
The industry may still have time before competitive pressure becomes acute, but the client-side shift is already under way. Wealth managers that continue to treat AI as a peripheral productivity tool risk missing the more important change: clients are becoming more informed, more sceptical and more capable of testing what they are told.
For an industry built on trust, advice and relationship depth, that is the more immediate challenge.