Part 5 - Agentic AI in wealth & asset management

Part 4 argued that discretionary management is becoming private banking's default operating model. This final part looks at the technology that actually makes running that model, at scale, for a genuinely broad client base, possible and closes the loop back to where this series started.

There's a moment in every technological wave when the conversation shifts from whether to adopt to how fast to lead. For wealth and asset management, that moment has arrived. Agentic AI has crossed from experimental to operationalv and the firms treating it as a strategic imperative rather than an IT project will define the next decade of the industry.

What makes agentic AI different

Generative AI gave the industry a productivity boost, faster drafting, better search, sharper summarization. Useful, but agentic AI is a different category of tool. Where generative AI responds to a prompt, agentic AI acts on an objective: it breaks a goal into steps, makes decisions, uses tools, and executes a workflow with minimal human intervention. An agentic system doesn't just answer "which clients are at risk of attrition?", it identifies them, drafts a personalized outreach brief for each one, flags the highest-risk cases to the relationship manager, and schedules the follow-up. All before the advisor opens their inbox.

The scale of adoption is already real

According to EY's 2025 survey of 100 wealth and asset managers, 95% of firms have already scaled AI adoption to multiple use cases, and 78% are actively exploring agentic architectures specifically. On the banking side, McKinsey's 2025 tracking found that 50 of the world's largest banks announced more than 160 agentic AI use cases in a single year. This isn't a pilot wave anymore. It's becoming infrastructure.

Where the value is showing up

The clearest early impact is on advisor time. KPMG's work with a top-10 investment manager used an agentic assistant to generate personalized meeting agendas from advisor profiles and historical notes, cutting preparation time roughly in half and saving an estimated 20,000 hours a year across the organization. On acquisition and onboarding, agentic tools are cutting manual prospecting time and materially speeding up KYC and AML processing by integrating data sources and flagging anomalies automatically. In portfolio management, autonomous agents are handling rebalancing, asset location, and tax-loss harvesting at a scale no human team can match, the same operational lift that makes discretionary management viable for a mass-affluent client base rather than just the top tier, as Part 4 described. And in compliance and back-office functions, agentic deployments are already showing strong early accuracy on document classification and meaningful reductions in case processing time.

The risks worth taking seriously

Autonomy cuts both ways. Agentic systems introduce real goal-misalignment risk, optimizing for a measurable proxy rather than the client's actual outcome and, in interconnected markets, the possibility that many systems acting on similar signals at once amplifies volatility rather than dampening it. Data privacy exposure grows sharply once an agent is empowered to access and act on sensitive client information. Regulatory frameworks weren't built with autonomous agents in mind, and the gap between what these systems can do and what compliance infrastructure can actually oversee is widening quickly, which is why both IBM and EY are pushing "compliance by design," building governance into the architecture from the outset rather than retrofitting it later. Legacy infrastructure adds a further constraint: many firms still run on systems that were never designed to plug into autonomous agents, and a well-documented shortage of specialist AI talent means adoption speed will vary sharply across the industry.

What this means for leadership

Three priorities stand out for anyone serious about this. Treat AI governance as a board-level issue, not a back-office one, agentic systems making consequential decisions on a client's behalf need oversight that's rigorous and continuously updated. Invest deliberately in the advisor-AI interface: the firms winning here aren't replacing advisors, they're redesigning what advisors spend their time on, freeing them for the relationship-intensive work machines can't replicate. And treat legacy architecture as a strategic constraint rather than an IT ticket — firms that can't plug agents cleanly into their core systems will find themselves structurally disadvantaged within three to five years.

Back to where this series started

Wealth management has long built its edge on information asymmetry, relationship depth, and portfolio construction skill. Agentic AI doesn't eliminate those advantages, but it commoditizes the operational infrastructure underneath them, fast and that's precisely what closes the loop on Part 1. Belgium's €300 billion in idle household cash isn't a product gap; it's a conversion gap that requires continuous, personalized, low-friction advisory delivered at a scale no advisory team can staff for manually. Agentic AI is the first technology that makes that kind of continuous engagement genuinely affordable across a bank's full client base, not just its wealthiest segment.

That's the throughline of this whole series: the €300 billion won't move because of a better product. It will move once banks combine a redesigned default (Part 1), an understanding of who they're now serving (Part 2), a clear-eyed view of what they're actually competing against (Part 3), an operating model built for scale (Part 4), and the technology that makes that model real (Part 5). That combination is exactly the kind of transformation we help clients at NORRIQ work through, not one piece in isolation, but the whole system.