Mercer is becoming Marsh.

Learn more
Learn more
Total rewards Transformation

The AI workforce paradox: Why people, skills and structures will determine success 

AI will do more of the work. Only people can make it pay off.

Insights from our joint Marsh and Oliver Wyman webinar in September, featuring a large Spanish bank.

Every bank and insurer has the same technology on the table: the same models, the same vendors, the same use cases. Yet the results could not be more different. A small group of firms are turning AI into measurable P&L impact; most are stuck in pilots.

On September 22, Marsh and Oliver Wyman jointly brought together HR and business leaders from across European financial services to explore why this is the case. We were joined by the Head of Strategy in Talent & Culture at a large Spanish bank. One message ran through the hour: The technology is ubiquitous; the humans are the differentiator.

A reality check: The prize is huge, the returns are not

Financial services is being reshaped on every front at once. GenAI use in banking jumped from 5% in 2023 to 45% in 2025, while regulation, legacy costs and new digital challengers squeeze margins. Yet while 88% of firms use AI, only around 6% see real bottom-line impact. Financial services leaders sense this: in our 2026 Global Talent Trends research, redesigning work to optimize AI and human capabilities ranked as the #1 leadership challenge in using AI for growth. Yet only 38% of FS organizations say they are able to adopt and embed new technology. Intent is near-universal; readiness is not.

That is the central paradox: AI will do more of the work, but only people can make it pay off. The real choice is not capability or cost; it is what you choose to scale. Scaling cost reduction alone delivers a one-off saving. Scaling capability with efficiency builds an advantage that compounds.

Much of how experts do their jobs is tacit knowledge, learned over years and never written down. For AI to learn it, people have to share it, which raises the question every employee is asking: why would I train the machine that might take my job? Without trust in leaders and a credible “new deal,” they won’t train AI, and the investment never pays back. The data suggests that trust is under real strain.
This chart is unable to display due to Privacy Settings.
The chart could not be loaded because the Privacy Settings are disabled. Under the "Manage Cookies" option in the footer, accept the “Functional cookies” and refresh the page to allow the chart to display.
Line chart comparing the percentage of employees who say they are thriving at work in financial services and across all industries. Both groups declined between the previous survey and 2026, with financial services falling by 26 percentage points.
Source: Marsh 2026 Global Talent Trends (all industries, N = 9,250) and Financial Services report
And leaders don’t see it. There is a 42-point gap between how well executives believe they understand employees’ needs and how employees see it. In one client example, the business units most confident about AI adoption scored up to 31 points higher on trust in leadership. Confidence in AI follows trust, not the other way around.
There is a huge hype around AI. But when we talk with our people, there is also anxiety, and we shouldn’t dismiss this concern.
Head of Strategy in Talent & Culture, a large Spanish bank

AI rarely removes whole roles. It breaks them into tasks that are automated, augmented or kept human-led. The real question is not how many roles will AI remove, but which tasks move to machines, and how do we help people move up the value chain? That redesign has a hidden cost: 33% of companies have already cut entry-level hiring. Remove the bottom rungs of the ladder and juniors can no longer learn by doing.
AI will not remove junior work. AI will redesign junior work.

AI agents are joining the org chart. A manager who today leads 8–10 people may soon also orchestrate 20–30 agents. Spans, layers, decision rights and new roles all need rethinking, yet only 54% of executives say they are prepared for human–machine teaming, down 14 points since 2024. Every people practice has to evolve, too, including the one most firms forget: reward. Rewarding output and tenure will undercut what AI needs most: judgment and knowledge-sharing. And the risks multiply quickly: 95% of firms have had at least one AI incident, and only around 2% report adequate guardrails.

From paradox to plan

Value builds as organizations move from individual productivity to redesigned processes and, ultimately, to new ways of working.

Productivity gains multiply as Al moves from tools to redesigned work

Indicative productivity uplift by horizon; most organisations are still in Horizon 1
5-20%

Horizon 1
Individual productivity

Employees use Al for everyday tasks

30-80%

Horizon 2
End-to-end process change

Workflows redesigned around agents

2-4x+

Horizon 3
Reimagined ways of working

Agents run processes; humans orchestrate

Source: Marsh 2026 AI Survey – Financial Services (three horizons; Oliver Wyman)
Most organizations get stuck in Horizon 1: people use AI, but the work itself doesn’t change. Moving beyond it means acting on three fronts at once.

Adoption doesn’t rise in a straight line. After the initial hype comes a dip of resistance and doubt, and this is where most AI programs stall. A clear narrative, managers equipped to listen, targeted upskilling and incentives that recognize AI-enabled performance carry people through.
If I started again, I would start earlier with managers. At the end of the day, managers are the most important people for change to stick.

Break roles into tasks, decide what to automate, augment or keep human-led, and redeploy the freed capacity deliberately to client contact, judgment and exceptions. In credit, the banker moves from processing applications to reviewing AI output and owning exceptions. In one agentic claims redesign, cycle times fell from 7–10 days to around 24 hours, at roughly 70–75% lower cost.

Spans widen, layers flatten and decisions move closer to the point of work, so decision rights and guardrails must become explicit.
We are working to industrialize agent development, but we need to industrialize control at the same pace.

Lessons from the frontline: A large Spanish bank

The bank, with more than 100,000 employees across Spain, Mexico, Turkey, Colombia and beyond, sees AI as a more profound transformation than the digital revolution. Its journey has run in three phases: adoption, giving every employee GenAI tools backed by training and a network of around 750 internal AI champions; test-and-learn on high-potential use cases; and now industrialization, led by an AI transformation unit reporting directly to the Executive Chairman.
Adoption has never been about providing the technology. The real objective is to help people use it in their day-to-day work, experiment and transform the way they work.
On trust, the bank listens to understand not just sentiment but readiness, asks managers to create safe spaces to experiment, and tracks adoption by country and business area. Employees can see their own usage, but managers deliberately cannot, so the data encourages experimentation rather than surveillance. For junior roles, it is creating new learning experiences, such as earlier exposure to clients and more time with senior bankers, because, as our guest put it, critical thinking is not something you can learn from a course. The biggest lesson was about scale:
Thousands of people experimenting does not mean that [the bank] is learning as an organization.

The real opportunity

AI is everywhere, and soon it will be a commodity. Advantage will come from the organizations that redesign themselves around it: those that build trust, rebuild work around skills and judgment, and reset the people systems that tell employees what matters. Return on investment depends on human engagement.

It is time to move from AI enthusiasm to AI economics.

About the author(s)
Sadiye Azisik

Financial Services Industry Leader, Marsh Europe & UK

Nick Starritt

Employee Experience Region Lead Consultant, Marsh Europe & UK

Lauren Lyons

Banking and Financial Services, Performance Transformation, Oliver Wyman

Related products for purchase
Related solutions
Related insights

Get in touch!

To discuss how we help firms assess workforce readiness, redesign work and skills, and build AI governance, please fill out the form.