Our Take Special Edition

A new risk framework for operational resilience

Hero image
  • Insight
  • 5 minute read
  • November 06, 2026

Why second- and third-line leaders need a more scenario specific time-sensitive and business-led framework for the next generation of cyber and operational risks

Key takeaways

  • An integrated framework for ops resilience and cybersecurity risks. AI-accelerated attacks and a widening range of hostile scenarios can turn a single weakness into a simultaneous cyber, data integrity and continuity crisis. Institutions need to manage cyber, business continuity and disaster recovery as one resilience discipline rather than as separate silos.
  • Trusted-state restoration of critical business services. Recovery must assume that production systems, recovery environments and backups may all be compromised. The objective therefore shifts from recovering individual applications to restoring an end-to-end business service, including correctly sequenced shared services, to a state the institution can trust.
  • A shared digital operational model across the three lines. Discovery, observability and dependency investments already under way in the first line can become a common foundation that links technology to business services, controls and evidence. Technology, Cyber, Risk and Audit can then ask different questions of the same facts while keeping their judgments independent.
  • Time- and scenario-specific risk appetite that directs investment. Clear tolerances for service disruption at defined time horizons, under severe-but-plausible scenarios, should set recovery objectives and investment priorities. The goal is risk-appropriate resilience that directs scarce resources to the most consequential exposures, not maximum resilience everywhere.
  • Evidence-based testing and assurance. Integrated, scenario-based testing, with second-line challenge and independent Audit assurance, shows whether critical services can stay within tolerance when dependencies fail. This aligns with the direction already set by the PRA and DORA and positions firms ahead of more prescriptive mandates.

Operational resilience can no longer be managed as separate silos of cyber, business continuity and disaster planning. AI is compressing attack timelines, and geopolitical conflict is widening the range of credible hostile scenarios, so all operational risk professionals across the three lines of defense face new demands. Financial institutions need a better framework that anchors recovery priorities and investment decisions to a clear risk appetite and gives every function, from the front office to Audit, a shared foundation of facts, scenarios and evidence from which to ask questions.

The risk is changing faster than the framework, making integration essential

For over two decades, operational risk frameworks at large financial institutions have helped bring discipline to risks that were once treated as a collection of unrelated control failures. Risk Control Self Assessments (RCSAs), operational risk loss tracking and lessons learned processes improved the visibility of risks. Those tools remain useful but are insufficient for a new reality of dynamic technology dependencies, time-dependent impact and a heightened need for demonstrated recoverability.

They were not necessarily designed for the class of high-impact technology disruptions that, while historically infrequent, are now growing fastest in terms of their contribution to risk. The underlying threat environment has shifted and what were once severe-but-unlikely “tail-risk” scenarios increasingly include credible threats to the survival of an institution. Frontier AI systems are becoming more capable at software exploitation, cyber-reconnaissance and autonomous action. A growing series of frontier-AI exploits, autonomous attacks and model “escapes” suggests that capabilities once treated as theoretical are beginning to appear in real-world environments, while sophisticated actors are also using AI to accelerate existing attack techniques.

For banks and other financial institutions, a rethink of the risk framework is necessary because technology disruption is increasingly a hybrid risk. A single weak point, such as a stolen credential, can turn a cyber incident into a data integrity and business continuity crisis at the same time. Attackers are also moving faster than the controls meant to stop them. Escalation chains built for incidents that unfold over hours fail when autonomous tools can break in and spread at machine speed.

From disaster recovery to trusted-state restoration

The hybrid nature of these threats changes the focus of recovery and resilience. Traditional disaster recovery often assumes that while the primary environment is unavailable, the recovery environment is trustworthy. A sophisticated cyber event can quickly invalidate that assumption. Production systems may continue operating even as credentials, code, data, or shared infrastructure are compromised. Disaster recovery environments may be exposed to the same threats through shared identity systems or administrative access, while backups may contain attacker persistence mechanisms, undermining confidence in the ability to recover safely.

A modern resilience framework must integrate capabilities that have often been governed separately: cyber detection and containment, forensics, determination of the last reliable or trusted state, immutable and isolated recovery copies, clean recovery environments, and validation before reconnection. The framework must also provide for shared services to be restored in the right sequence. Recovering a payments or servicing application is of little value if Active Directory, DNS, Public Key Infrastructure, Kafka, Kubernetes, network services or key APIs remain unavailable or untrusted.

The framework’s main objective is to restore not just applications but a viable business service to a state the institution can trust. This distinction is central to risk management because the right control architecture for a two-minute outage may be completely different from the architecture needed for a two-day or two-week disruption. Because these disruption events also rapidly constrain the resources available to respond (i.e., both personnel and infrastructure), the framework must provide the flexibility to make deliberate tradeoffs among competing priorities in order to protect the viability of the institution as a whole. Some services may require near-continuous operation with all business capabilities available; others can degrade safely for a period; still others may need to be suspended while the firm protects customers, counterparties and markets and rebuilds capability.

The control question is changing

Is the application recoverable? → Can the critical business service be restored end-to-end? 
Is there a disaster recovery site or backup? → Can the firm recover from a trusted state if primary and recovery environments are compromised?
Is the control in place?  → Does evidence show the control works under the relevant scenario and time horizon?

From technology inventory to risk intelligence

Answering these new questions requires better, more up-to-date information about the institution and how it operates. Much of the necessary investment is already happening in the first line. Technology, Cyber and Operations functions are combining asset inventories with discovery, observability and dependency tools to understand what should exist, what is running, and how they are all inter-connected. As technology, APIs, cloud services and third-party relationships change faster than periodic inventories can be updated, Risk and Audit also need better visibility to properly identify and assess material exposures and controls. 

The opportunity is to build a common model of the institution that serves all three lines. Technology, Cyber, Risk and Audit do not need their own parallel technology inventories. They need a common factual substrate connecting technology dependencies to business services, controls and evidence. Technology may ask whether a dependency is operating correctly; Cyber whether it creates an attack path; Risk whether it creates an unacceptable concentration; and Audit whether the evidence supports management’s conclusions. The facts can increasingly be shared even when responsibilities and judgments remain appropriately independent.

That makes design and timing important. As firms continue to build these capabilities in the first line, they should consider the information Risk and Audit will eventually need: common identifiers and taxonomies, data provenance and history, links to critical business services and controls, and evidence that can support risk appetite, scenario analysis and assurance. The same foundation, tailored to augment the existing capabilities of CMDBs and GRC tools, can better connect RCSAs, operational-risk capital and enterprise stress scenarios as well as provide Audit a more dynamic view of its auditable universe.

Over time, this shared substrate can evolve into a true digital operational model of the firm. Knowledge graphs are a promising tool to organize this information because resilience risk is inherently relational: applications share infrastructure, services depend on multiple technologies and third parties, and the importance of those relationships changes by scenario and duration. Connecting business impact, architecture, controls, dependencies, recovery evidence and observed activity can enable analysis of blast radius, concentrations, recovery sequencing and identify where additional investment reduces risk most. The value of these models, however, will depend on disciplined reconciliation of authoritative records with observed infrastructure, clear data provenance and explicit recognition of gaps and uncertainty; a more sophisticated model will not necessarily provide a more accurate view of risk. 

The direction of how this architecture will evolve is increasingly clear. Better discovery, cheaper computing and AI are making operational resilience risk models more up to date, more connected, and easier to interrogate. As with early financial-risk models, they need not represent everything to create value. Firms can start with critical business services and consequential scenarios, then improve the data and expand the model over time – one increasingly reliable representation of the operating environment, with different questions asked by each line of defense.

Operational resilience needs its financial-risk moment

Operational resilience may be approaching the moment market risk reached decades ago. Early variance-covariance models were crude compared to today’s cross-asset, nonlinear risk engines, but they were useful: they let firms compare exposures on a common basis and connect individual trading decisions to enterprise risk before every interaction could be modeled precisely.

Financial-risk modeling then improved cumulatively as computing became cheaper, data improved and models incorporated more products, correlations, counterparty effects, collateral, stress scenarios and time horizons. Rapid advances in threat detection, discovery, observability and AI, combined with a growing convergence around common resilience challenges, could now accelerate a similar transformation by attracting both established technology providers and new entrants to develop increasingly sophisticated solutions.

Operational resilience can follow the same path: start with a model useful enough to answer consequential questions now and make the information architecture richer as technology and analytical capabilities improve.

Risk appetite must become time- and scenario-specific

To effectively assess potential risk events, business and risk managers need more than simply knowing that an application has a four-hour recovery time objective (RTO). Rather, they need information on how harm to customers, counterparties, markets and the institution compounds over time, and which business outcomes become intolerable first. Losing access to a critical market utility overnight could create a very different threat to the firm than losing a customer-facing capability for the same period. Some services can be designed for graceful degradation; others may need independent capacity or a credible manual path; some may require the ability to stop safely and rebuild.

Most risk appetite statements do not yet support these decisions. A generic statement that the firm has a low appetite for disruption does not say which services to protect first, for how long, or against which scenarios. Institutions need to establish clear appetite for service disruption at defined time horizons, under defined severe-but-plausible scenarios. Those limits then become the test for the design. Recovery objectives should follow from what the firm can tolerate, not from what the current architecture happens to deliver.

This exercise will point to required investment. Institutions may find that they need more redundant capacity, stronger segregation, isolated recovery environments, independent backup services, additional on-premises capability, multiple cloud regions or providers, greater automation, and more frequent end-to-end testing to stay within appetite. Diversification should not become a slogan. Multi-region architecture does not solve a prolonged provider-wide outage; infrastructure-as-code does not help if the control plane needed to deploy it is inaccessible; multi-cloud can itself create complexity and new failure modes. The question is which combination of diversity, redundancy, segregation and substitution materially reduces the scenario-specific risk.

That cost is exactly why the risk framework matters. The objective cannot be maximum resilience everywhere. It must be risk-appropriate resilience: directing scarce engineering capacity and resilience investment to the exposures that can most damage the institution or its critical businesses.

Regulatory direction is reinforcing the shift

Regulation will tend to reinforce this shift, but likely at different speeds across regions. The UK’s PRA framework anchors on important business services, impact tolerances and testing against severe-but-plausible disruption; the EU’s Digital Operational Resilience Act brings information and communications technology (ICT) risk, incident response and third-party technology risk into one framework. Both move the question from whether controls exist to whether critical services stay within tolerance when dependencies fail. For globally active institutions, understanding these relationships is already an expectation, not a future one, but regulation does not hand firms a ready-made digital model of the institution. The harder task, and the real opportunity, is building the connected information and control architecture regulators are pointing toward before it is mandated in more prescriptive form.

Rethinking resilience testing across the three lines of defense

This evolution also has implications for how resilience is tested and independently assured. Traditional control testing and reviews of individual risk areas may not fully demonstrate whether critical business services can withstand complex disruptions. First-line testing needs to move toward more integrated, scenario-based approaches that validate how controls, processes and recovery capabilities work together. The second line plays an important role in challenging testing approaches and results, while Audit provides independent assurance over the effectiveness of the overall resilience framework, the completeness of the universe of systems, data, applications and connectivity and the reliability of management’s conclusions. A stronger framework would challenge whether the firm can answer six questions with evidence:

1. Business impact and time Which services can the firm not afford to lose for two minutes, two hours, two days or two weeks, and how does customer, market, counterparty and franchise harm compound over those horizons?
2. Scenario specificity Which severe-but-plausible scenarios matter for each critical business service, including hostile compromise, trusted-state loss, regional or provider outage, dependency failure and insider or social-engineering attack?
3. Concentration and common mode Where do individually resilient applications converge on the same identity, network, cloud, data, middleware, provider or counterparty dependency?
4. Recovery strategy For each scenario, should the service absorb the event, fail over, degrade gracefully, substitute capacity, recover from a trusted state or stop and rebuild?
5. Demonstrated capability What evidence shows the business service – not merely its applications – can operate or recover within tolerance, including shared-service sequencing and third-party dependencies?
6. Investment efficiency Which additional control, redundancy or diversification investment produces the greatest reduction in residual risk, and which expensive measures add little protection against the scenarios that matter?

A framework oriented around these questions better equips senior management and boards to understand where the institution sits relative to appetite. A portfolio view of the services and dependencies that could threaten major business lines or the institution, the scenarios under which current resilience could fail, the time available before impacts become intolerable, and the cost and effectiveness of further mitigation provides the critical inputs for informed decisions about the organization’s resilience posture.

The goal is risk-appropriate resilience: knowing what the firm must preserve, for how long, against which scenarios, and where the next dollar of resilience investment reduces the most consequential risk.

Operational resilience has outgrown the frameworks built to manage it. Threats can now move at machine speed, compromise the environments firms rely on to recover, and turn a single weakness into a cyber, data and continuity event at once. Meeting them requires a different approach. Firms need to be able to restore critical business services to a trusted state rather than recover applications alone, set clear risk appetite as tolerances for defined time horizons and scenarios, and test whether the evidence shows those tolerances can be met. Because resilience is costly, they also need to direct investment where it removes the most risk, aiming for risk-appropriate resilience rather than maximum resilience everywhere. 

Much of the information needed is already being assembled in the first line; the opportunity is to bring these threads together while the technology foundation is being built. If Risk and Audit help shape that foundation now, one set of facts – a progressively richer digital model of the firm – can support operations, enterprise risk decisions and independent assurance while laying the groundwork for a more resilient institution.

Operational resilience has outgrown the traditional operational risk framework

Authors

Stein Berre

Managing Director, PwC US

Shawn Lonergan

Cyber, Data, and Tech Risk, Principal, PwC US

Follow us