By Salma Datenis, Vice President and Head of Cloud Studio at Amdocs. Amdocs were finalists in ‘Best FinTech for Corporate Banking / Lending and Brokerage’ award at the 2026 FinTech Awards.

For years, the cloud playbook in financial services looked straightforward. You moved more workloads, automated more infrastructure, and standardized operations. The approach produced real gains, and for a while, the trajectory felt clear.

Yet many banking and insurance leaders are now reaching a plateau. After years of heavy automation investment, cloud operations remain slower, costlier, and more cumbersome than they should be, while the gap between expectation and reality keeps widening.

The teams who should be driving strategic change are instead coordinating across systems, chasing approvals, and managing exceptions. Complexity, meanwhile, compounds faster than anyone can resolve it. And the gap between where teams spend their time and where the organization needs them shows up in every cloud program that runs slower than it should.

Cloud operations stall at the point where human judgment is required

Traditional automation handles repeatable tasks well. But it struggles in environments where conditions shift constantly; dependencies are hidden, and exceptions are the norm. Modern cloud estates, spanning multiple clouds, hybrid infrastructure, and legacy systems, are precisely that type of environment. It’s where every change ripples across the environment, every optimization involves trade-offs, and every operational issue arrives with context that must be understood before anyone can act.

The real source of delay, though, is the human work that automation can’t replace, including interpreting policies, mapping dependencies, securing approvals, reconciling exceptions, and deciding what to do next.

In financial services, this friction creates consequences that go far beyond operational inconvenience. A performance issue can become a resilience issue; a migration decision can create compliance exposure, and a cost action can affect third-party relationships or service continuity. So, the cloud conversation changes from a technical discussion to a business-critical one, raising the stakes of every decision that stalls.

This is why an organization can appear highly automated on paper and still feel slow in practice.

More automation can make the problem worse

The instinctive response is to automate more. Yet while scripts and pipelines execute rules effectively, they can’t interpret ambiguity, reconcile conflicting priorities, or adapt when conditions shift underneath them. And as environments grow more complex, adding more automation increases cognitive burden, with more systems to monitor, more alerts to triage, more points where human intervention is still required.

The result is an organization that has invested heavily in automation but still slows down precisely where it matters most. The missing ingredient is judgment and decision-making capability; something automation alone will never provide.

From task execution to decision execution

The more important shift now underway is about changing what gets automated. Rather than executing tasks, the next generation of cloud operations needs to orchestrate decisions – interpreting context, weighing dependencies, handling exceptions, and acting within defined boundaries.

At Amdocs, we refer to this as Agentic Cloud, a cloud operating model in which AI agents orchestrate decisions and actions across increasingly complex environments, operating within enterprise guardrails and human oversight. The goal is significantly less manual coordination, freeing people to focus on architecture, resilience, and strategic change.

Where the real value sits

Deployment of AI agents is a generational change for operations. However, the costs of agents cannot be ignored. Therefore, they should be applied with discretion, in selected “high-impact zones” that will deliver an ROI. Selecting the applicable challenges is a science, but examples would include decision-dense workflows where governed autonomy matters:

• Modernization and migration, where application discovery, dependency mapping, and sequencing delay progress
• Cloud operations, where teams have dashboards and alerts but still struggle to turn insight into action
• Cost optimization, where inefficiencies are visible, but decisions are delayed by fragmented ownership and a lack of expertise

• Governance-heavy workflows, where every change must be explainable, auditable, and reversible.

Autonomy without governance isn’t an option in financial services

Financial institutions are often hesitant to adopt new technologies due to concerns around financial, reputational or regulatory impacts. So, it is only natural to view all new technologies with a degree of caution.

Part of the caution stems from a misunderstanding of how agentic AI works.

AI agents don’t operate at a single fixed level of autonomy. In practice, current frameworks generally describe a spectrum, from agents that assist and recommend through to agents that act independently within defined boundaries, with most deployments today sitting toward the lower end of that spectrum. For most institutions, that’s where the journey begins.

The longer-term trajectory, however, is toward greater operational autonomy. In this context, autonomy is never unsupervised. What makes agentic AI viable in regulated environments is governed autonomy, where agentic execution is policy-aware, observable, auditable, explainable, and constrained by enterprise intent, with human oversight maintained at every critical stage.

Indeed, the governance model matters as much as technology itself. When governance is bolted on after the fact, adoption stalls because the controls feel like obstacles rather than enablers. So when it’s embedded directly into how work runs, through role-based permissions, approval boundaries, and full traceability, autonomy becomes something an institution can actually use. In financial services, that distinction is everything.

Organizations that act decisively will define the model on their own terms

The market is already moving, and the pressure driving it is real. Financial institutions are managing simultaneous demands to modernize faster, control more tightly, reduce friction without increasing risk, and improve resilience while keeping costs in check. These priorities have to be addressed together, and the current cloud operating model was not designed for this.

According to a recent Amdocs’ 2026 research, 71% of enterprises expect to run AI agents in cloud environments by 2026, and 68% already view agentic AI in cloud as a competitive advantage. What began as experimentation is becoming part of mainstream enterprise planning, and the institutions that move early and thoughtfully have more room to shape the model on their own terms.

Decision velocity is the new competitive measure

Agentic Cloud represents a fundamental shift in how cloud operations are run. But the winners here won’t be the ones that pilot more AI. They will be the ones that redesign workflows, strengthen data readiness, clarify decision rights, and modernize governance alongside technology.

Moving from signal to action quickly and confidently, what Amdocs calls decision velocity under governance, is becoming the defining measure of cloud maturity. Once embedded early, this capability is significantly harder to replicate than any technological deployment.

A final thought for FSI leaders

For years, cloud leadership was measured by migration progress, platform maturity, and automation coverage. Those metrics still matter. Yet financial services institutions are moving into a new phase that’s defined by decision complexity and operational efficiency rather than automation coverage alone. The institutions that recognize this first will operate with less friction, more control, and a far stronger position to manage the complexity still ahead.