Can wealth managers embrace AI while managing the risks?
4-minute read . 24 Oct 2025
Here's the reality from a European Financial Services AI Pulse Survey run by EY and FT Longitude. 410 leaders across banking, insurance, and wealth and asset management (including Luxembourg) responded between March and June 2025. The message: investment is high, confidence on risk is shaky, and agentic AI is arriving faster than comfort levels.
Investment is up. Risk readiness lags.
Firms are spending big on the foundations. 88% report moderate to extensive investment in AI training, 84% in model testing and audits, and 83% in data access control.
Still, 57% say their tech risk approach isn't fit for emerging AI. That jumps to 60% for wealth and asset managers. A worrying 30% have no or limited controls to reduce bias. Only 52% lean on internal audit for assurance, with others turning to expert consultation and third-party validation. Banking and capital markets report the strongest controls, but even "Transforming" and "Leading" firms admit gaps.
Agentic AI is in play, even with low comfort
Over 40% of financial services firms report using agentic AI today. Wealth managers are even more active, with usage already above 40%, despite only a third feeling genuinely comfortable with it. Another 25% plan adoption within six months.
Familiarity with other AI capabilities (multimodal, synthetic data, quantum machine learning, autonomous robots) stays below 50% for wealth managers. Interestingly, autonomous robots are expected to see wider adoption in the next year. Many see these as stepping stones before agentic AI orchestrates complex workflows across the firm.
Leaders fear job losses, manipulation, and duller work
Executives worry about job cuts, manipulated perceptions, and false content. Wealth and asset managers are the most concerned across subsectors. Only 32% believe consumers trust their sector to use AI in their best interest.
Accountability, transparency, data protection, cybersecurity, ethics, and disinformation sit high on the list. With strict regulations and sensitive data, the reputational hit from opaque or biased AI-driven decisions can be severe.
How to embrace AI while managing risk
1) Tie AI to clear business outcomes
- Prioritize use cases with measurable upside: faster investment research, improved personalization, better client servicing, KYC/AML triage, operations automation.
- Define success upfront: cost-to-serve, alpha support, client NPS, cycle time, error rates, risk events, and audit findings.
- Build the data base layer: governed sources, lineage, permissions, encryption, retention, consent management, and quality thresholds per model.
- Set decision rights: who sponsors, who approves, who monitors, and when to pause or roll back.
2) Evolve risk and governance as AI scales
- Create an AI governance committee across risk, compliance, IT, data, legal, and investment teams. Give it real authority.
- Maintain an AI use-case inventory with model cards, owners, data sources, risks, controls, and sign-offs.
- Adopt proven references like the NIST AI Risk Management Framework and prepare for the EU AI Act.
- Test before release: bias and fairness checks, explainability, adversarial testing, red-teaming, data leakage scans, and prompt/content safety reviews.
- Keep humans in control for client-facing and high-impact decisions. Write down escalation rules and thresholds.
- Monitor in production: drift, performance, hallucination/error rates, access logs, and incident flags. Add kill switches.
- Use independent assurance: internal audit, model risk, and third-party validation. Report to the board quarterly.
- Manage vendors like models: due diligence, SLAs, security testing, and exit plans.
3) Bring your people with you
- Train advisors, portfolio managers, compliance, and risk teams on practical AI use and limits. Make it role-specific.
- Redesign workflows so AI assists, not replaces. Free experts to focus on strategy, judgment calls, and clients.
- Be open with clients about how AI is used, where humans review, and what benefits they should expect.
- Set "agent" guardrails: sandbox environments, scoped permissions, rate limits, audit logs, and clear ownership.
What the numbers say about value
Across sectors, AI is already linked to higher profits or lower costs. Gains are strongest in advanced manufacturing, with public sector and healthcare moving slower. Nearly half of employees report productivity improvements, with managers seeing the biggest lift.
To make this real, measurement needs an upgrade. Instrument workflows to capture near real-time productivity, quality, risk, and client outcomes. That reduces reporting gaps between leadership and teams and speeds up decisions.
90-day execution plan for wealth managers
- Publish a one-page AI mandate tied to revenue, cost, and risk outcomes.
- Stand up an AI governance committee with weekly rhythms and a clear charter.
- Create a live AI use-case and model inventory. Add owners and KPIs.
- Pick three high-value pilots (e.g., research summarization, client reporting, KYC triage) and define guardrails.
- Run bias, explainability, and adversarial tests before going live.
- Instrument production: logging, alerts, drift checks, and fail-safe rules.
- Launch targeted training for frontline teams and model owners.
- Brief the board on risks, controls, and milestones. Agree on decision gates.
Close the skills gap
The biggest constraint is AI fluency across the business. Practical training accelerates adoption and reduces risk. If you're setting up role-specific enablement, explore curated learning paths by role and function.
The takeaway is simple: invest with intent, raise the bar on controls, and keep skilled people in the loop. Do those three, and AI becomes an operational advantage, not a liability.
Your membership also unlocks:
