The Capability Gap Is Structural, Not Cyclical

Regional and community banks have always competed on relationship depth rather than product breadth. For decades, that was a defensible position. Local knowledge, faster decisions, and genuine accountability to the communities they served gave smaller institutions a meaningful edge over the impersonal scale of national banks. That edge has not disappeared, but it is narrowing in ways that should concern every regional bank board.

The issue is not that national banks have become better at relationships. It is that technology has begun to replicate the speed and personalisation that once required human judgment. AI-enabled underwriting can process a small-business loan application in minutes. Digital servicing platforms can anticipate a customer's next financial need before the customer articulates it. These are not marginal improvements — they are structural shifts in what customers expect from any bank, regardless of size.

Where AI Creates Genuine Opportunity

The instinct among many regional bank leaders is to treat AI as a cost-reduction tool — a way to automate back-office processes and reduce headcount in operations. That instinct is understandable but strategically limited. The more consequential opportunity lies in using AI to extend the relationship model that regional banks already do well.

AI-assisted underwriting, for instance, does not replace a loan officer's judgment — it sharpens it. A model that surfaces relevant risk signals and comparable credit histories allows a skilled officer to make a better decision faster. Similarly, AI-driven deposit analytics can help relationship managers identify which customers are at risk of moving balances before the conversation becomes reactive. These applications amplify human capability rather than substitute for it, which is precisely the kind of AI deployment that fits the regional bank model.

The Data Foundation Problem

The practical obstacle for most regional banks is not access to AI tools — it is the state of their underlying data. Core banking systems that were implemented a generation ago were not designed to support the kind of real-time, customer-level analytics that modern AI requires. Data sits in silos across lending, deposits, and wealth management, often in formats that cannot be easily reconciled.

Addressing this does not require a wholesale core replacement, which is expensive and disruptive. It requires a deliberate data architecture investment — a layer that aggregates and normalises customer data across systems without replacing them. Banks that make this investment now will be positioned to deploy AI applications progressively as the tools mature. Banks that defer it will find themselves unable to act when the competitive pressure becomes acute.

A Credible Path to Parity

The strategic question for regional bank leadership is not whether to invest in AI — that decision has effectively been made by the competitive environment. The question is how to sequence the investment in a way that builds capability without overwhelming an organisation that is not structured for large-scale technology transformation.

The most effective approach we have observed starts with two or three high-value use cases where AI can demonstrably improve a decision that matters to customers — underwriting speed, deposit retention, or fraud detection. Proving value in a contained domain builds internal confidence, develops the data infrastructure that broader deployment will require, and creates a template for governance that regulators will expect to see. Parity with national institutions is achievable, but it requires treating AI as a strategic priority rather than an IT project.

Why Piecemeal Reform Fails

Health systems have spent the better part of two decades attempting to improve performance through sequential initiatives — a patient access programme here, an electronic health record implementation there, a workforce redesign effort that runs in parallel with neither. The results have been consistently disappointing, not because the individual initiatives were poorly conceived, but because the problems they address are deeply interdependent.

A health system that improves scheduling technology without addressing the staffing model that determines who answers the phone will not see the access improvements it expects. A workforce redesign that does not account for the technology constraints clinicians work within will generate resistance rather than adoption. The organisations that have achieved durable operational improvement have done so by treating access, staffing, and technology as a single integrated problem rather than three separate workstreams.

Redesigning Patient Access as a System

Patient access is the front door of a health system, and most front doors are poorly designed. Scheduling logic built around provider convenience rather than patient need, referral pathways that require multiple handoffs before an appointment is confirmed, and digital tools that replicate paper processes rather than reimagining them — these are the access failures that drive patients to competitors and erode the loyalty that health systems depend on.

Effective access redesign starts with mapping the patient journey from the moment a need is recognised to the moment care is delivered, and identifying every point where the system creates friction. The goal is not to automate the existing process but to eliminate the steps that exist only because of legacy constraints. Health systems that have done this work have reduced time-to-appointment significantly and improved patient satisfaction without adding capacity.

Staffing Models Built for Flexibility

The staffing crisis in healthcare is real, but it is not uniform. The organisations struggling most acutely are those with rigid staffing models that cannot flex with demand — fixed ratios, narrow role definitions, and scheduling practices that treat every shift as identical. The organisations managing best have built staffing models that distinguish between tasks that require clinical judgment and tasks that can be performed by differently trained staff, and have invested in the cross-training that makes flexible deployment possible.

This is not a cost-reduction argument. It is an argument for designing staffing around patient need rather than historical convention. A model that deploys the right skill level for each task is both more effective and more sustainable than one that defaults to the highest-credentialed available person for every interaction.

Technology as Enabler, Not Solution

Technology investment in health systems has consistently underdelivered because it has been treated as the solution rather than the enabler. Electronic health records were supposed to improve care coordination; in many systems they have increased administrative burden without improving clinical decision-making. The problem is not the technology — it is the assumption that deploying technology changes behaviour without changing the processes and incentives that drive behaviour.

Health systems that get technology right invest as much in implementation design as in the technology itself. They define the workflow changes that the technology is meant to support before they configure the system, and they measure adoption and behaviour change rather than go-live milestones. That discipline is unglamorous, but it is the difference between a technology investment that transforms operations and one that adds cost without changing outcomes.

The Speed Trap

The dominant narrative in enterprise AI is one of urgency. Move fast, experiment broadly, and accept that governance will follow capability. This framing is understandable — the competitive pressure to demonstrate AI progress is real, and the tools are genuinely powerful. But the organisations that have moved fastest without building governance foundations are already encountering the consequences: model outputs that cannot be explained to regulators, decisions that cannot be audited, and liability exposure that was not anticipated when the use case was approved.

Speed and responsibility are not opposites. The firms that will sustain AI advantage are those that build accountability into their adoption model from the beginning, not as a constraint on progress but as the infrastructure that makes progress durable.

What Governance Actually Requires

AI governance is frequently misunderstood as a compliance function — a set of policies and review processes that sit above the work and slow it down. Effective governance is something different: it is the set of decisions, accountabilities, and standards that allow an organisation to deploy AI at scale without creating risks it cannot manage.

In practice, this means three things. First, a clear taxonomy of AI use cases by risk level, so that the review process is proportionate — a customer-facing credit decision requires different oversight than an internal document summarisation tool. Second, defined ownership for each deployed model, including accountability for monitoring performance and responding when the model behaves unexpectedly. Third, a feedback loop between the teams deploying AI and the teams managing risk, so that governance evolves as the technology does rather than lagging it by two years.

The Trust Dividend

Organisations that build responsible AI practices earn a trust dividend that compounds over time. Customers who understand how AI is being used in decisions that affect them are more likely to accept those decisions, even when the outcome is not in their favour. Regulators who can see a coherent governance framework are more likely to engage constructively rather than prescriptively. Employees who trust that AI is being deployed thoughtfully are more likely to adopt it and less likely to work around it.

This is not a soft argument. Trust is a competitive asset, and in industries where AI is being used to make consequential decisions — credit, insurance, healthcare, employment — the organisations that can demonstrate trustworthy AI will have a structural advantage over those that cannot.

Building the Capability

Responsible AI adoption requires capabilities that most organisations do not yet have at scale: the ability to explain model outputs in plain language, to detect and correct for bias in training data, to monitor model performance over time, and to retire models that are no longer fit for purpose. Building these capabilities is not primarily a technology problem — it is an organisational design problem.

The organisations doing this well have created cross-functional AI governance teams that include legal, risk, technology, and business representatives, and have given those teams genuine authority to pause or modify deployments. They have also invested in training that helps business leaders understand what AI can and cannot do, so that use-case decisions are made with realistic expectations. That combination of structure and literacy is what separates responsible AI adoption from responsible AI theatre.

The Cost of Optimised Fragility

For two decades, manufacturing strategy was dominated by a single imperative: eliminate waste. Lean principles, just-in-time inventory, and single-source supplier relationships were the tools of choice, and they delivered genuine results — lower working capital, higher asset utilisation, and cost structures that made global competition viable. The assumption embedded in all of it was that the supply environment was stable enough to be optimised.

That assumption has been tested repeatedly and found wanting. The manufacturers that entered recent disruptions with the leanest networks suffered the most acute shortages and the longest recovery times. The problem was not that lean principles are wrong — it is that they were applied without adequate consideration of the tail risks they create. A network optimised for efficiency in normal conditions is a network that fails catastrophically in abnormal ones.

Deliberate Redundancy as Strategy

Resilience requires accepting some inefficiency as the price of stability. This is a difficult argument to make in organisations where cost discipline is deeply embedded, but the framing matters. The question is not whether redundancy costs money — it does. The question is whether the cost of redundancy is less than the expected cost of the disruptions it prevents.

For most manufacturers, a rigorous analysis of that trade-off reveals that targeted redundancy in critical components and sub-assemblies is clearly justified. The challenge is identifying which nodes in the network are genuinely critical — where a single failure would halt production — and building alternative sources or buffer inventory specifically for those nodes. This is not a blanket argument for carrying more inventory; it is an argument for carrying the right inventory in the right places.

The Multi-Tier Visibility Problem

Most manufacturers have reasonable visibility into their tier-one suppliers. They know who they are, what they supply, and roughly how financially stable they are. The problem is that the risks that cause the most severe disruptions typically originate further up the chain — at tier two or tier three, where a single specialised component manufacturer supplies multiple tier-one suppliers simultaneously.

Achieving multi-tier visibility requires a different kind of supplier relationship than most manufacturers currently maintain. It means investing in the data infrastructure to map the extended supply chain, and it means working with tier-one suppliers to understand their own dependencies. This is time-consuming and requires supplier cooperation that is not always forthcoming. But the manufacturers that have built this visibility have consistently been able to anticipate disruptions that their competitors discovered only when production stopped.

Resilience as Competitive Advantage

The manufacturers that have invested in network resilience are beginning to discover that it is not merely a defensive capability — it is a source of competitive advantage. When a disruption hits an industry, the manufacturers that can continue to supply customers while competitors are scrambling earn loyalty and market share that persists long after the disruption ends.

The strategic implication is that resilience investment should be evaluated not only on its cost but on its revenue protection and growth potential. A manufacturer that can credibly promise supply continuity to key customers in a disrupted environment has a differentiated value proposition. Building that capability requires upfront investment, but the return — measured in customer retention and share gain during the inevitable next disruption — is substantial.

The Allocation Problem

Utilities face an investment environment unlike anything in their recent history. The combination of decarbonisation commitments, electrification demand growth, and ageing infrastructure creates a capital requirement that exceeds what most balance sheets can absorb without difficult trade-offs. The question is not whether to invest — the regulatory and competitive environment has answered that — but where to invest first, and how to sequence decisions that are deeply interdependent.

Most utilities are not yet making these trade-offs explicitly. Investment decisions are being driven by regulatory requirements, technology availability, and internal advocacy rather than by a coherent strategic framework that weighs the relative returns and risks of generation, grid, and storage investment. The result is a portfolio of commitments that may individually be defensible but collectively do not add up to a coherent strategy.

Generation: The Renewable Build-Out

The case for renewable generation investment is well established, and most utilities have committed to significant build-out programmes. The strategic question is no longer whether to build renewables but how to manage the transition risk — the period during which renewable capacity is being added but dispatchable generation is still needed to maintain reliability.

Utilities that are managing this transition well are maintaining a clear view of their reliability obligations and making retirement decisions for legacy generation assets based on when renewable and storage capacity will be sufficient to replace them, not on political or financial pressure to move faster. The utilities that are struggling are those that have committed to retirement timelines that their renewable build programmes cannot support.

Grid: The Underinvested Bottleneck

The grid is the most underinvested part of the energy system, and it is increasingly the binding constraint on the energy transition. Renewable generation that cannot be connected to load centres because transmission capacity is insufficient does not reduce emissions — it creates stranded assets. Distribution infrastructure that was designed for one-way power flow from central generation cannot efficiently manage the two-way flows that distributed generation and electric vehicles require.

Grid investment is less visible than generation investment and less amenable to the kind of project-by-project financial analysis that utilities are comfortable with. But the utilities that treat grid modernisation as a strategic priority rather than a maintenance obligation will have a structural advantage as electrification demand grows. The grid is the platform on which every other energy transition investment depends.

Storage: The Enabling Technology

Storage is the technology that makes the renewable-dominated grid viable, and its economics are improving rapidly. The strategic question for utilities is not whether storage will be important — it will — but how to build storage investment into a capital plan that is already stretched.

The utilities making the best storage decisions are those that have defined the specific reliability and flexibility problems they are trying to solve before selecting storage technologies and siting locations. Storage deployed to solve a specific grid constraint delivers a calculable return. Storage deployed because it is strategically important without a defined use case is capital at risk. The discipline to connect storage investment to specific operational problems is what separates effective energy transition strategy from well-intentioned capital misallocation.

The Fragmentation Problem

Most retailers know a great deal about their customers in aggregate and very little about them as individuals. Transaction data sits in the point-of-sale system. Loyalty programme data sits in a separate platform. Online browsing behaviour is captured in a third system. Customer service interactions are logged somewhere else entirely. Each of these data sources tells a partial story, and the partial stories are rarely reconciled into a coherent view of who the customer is and what they need.

The consequence is a customer experience that feels impersonal despite the volume of data being collected. Promotions that are irrelevant to the customer's actual purchase history. Recommendations that reflect what was popular last season rather than what this customer is likely to want. Service interactions that require the customer to re-explain their situation because the agent cannot see what happened in the last interaction. These failures are not the result of insufficient data — they are the result of data that is not connected.

Building the Identity Layer

The foundation of personalisation at scale is a unified customer identity — a persistent record that links every interaction a customer has with the brand across channels and over time. Building this requires resolving the identity problem: matching records across systems that use different identifiers, handling customers who interact both online and in-store, and managing the privacy requirements that govern how customer data can be used.

This is primarily a data engineering challenge, not a technology procurement challenge. The tools to build a customer data platform are widely available; the difficulty is in the data quality work that must precede deployment. Retailers that have invested in cleaning and reconciling their underlying data before building the identity layer have seen dramatically better results than those that have deployed the platform first and hoped the data quality problems would resolve themselves.

Personalisation That Earns Its Cost

Personalisation is not valuable in itself — it is valuable when it changes customer behaviour in ways that improve margin. The retailers that are generating real returns from personalisation investment are those that have defined specific behaviours they are trying to influence — basket size, visit frequency, category penetration — and have designed their personalisation programmes around moving those metrics rather than maximising the sophistication of the recommendation engine.

This requires a discipline that is harder than it sounds. The temptation in personalisation programmes is to optimise for engagement metrics — click-through rates, open rates, time on site — that are easy to measure but loosely connected to margin. The retailers that have resisted this temptation and maintained a clear line of sight from personalisation investment to margin improvement are the ones generating returns that justify continued investment.

The Margin Opportunity

The margin opportunity from effective personalisation is substantial and comes from two sources. The first is promotional efficiency — the ability to offer the right promotion to the right customer rather than blanket discounting that erodes margin across the entire customer base. The second is assortment optimisation — using customer-level data to make better decisions about what to stock, where, and in what quantities.

Retailers that have built the data foundation and the analytical capability to act on it are finding that personalisation is one of the highest-return investments available to them. The barrier is not the technology — it is the organisational commitment to treat customer data as a strategic asset and invest in the infrastructure and capability required to use it well.

The End of the Easy Returns

Private equity returns over the past decade were driven by a combination of factors that are no longer reliably available: falling interest rates that expanded multiples, abundant leverage at low cost, and a buoyant exit environment that rewarded almost any business with a credible growth story. Sponsors that generated strong returns during this period were not necessarily doing anything wrong — they were operating in an environment that was structurally favourable to financial engineering.

That environment has changed. Higher rates have compressed multiples and increased the cost of leverage. Exit markets are more selective. The businesses that command premium valuations today are those that can demonstrate genuine operational improvement — revenue growth that comes from better products and better go-to-market execution, margin improvement that comes from structural cost reduction rather than one-time savings, and management teams that have built the capabilities to sustain performance after the sponsor exits.

What Operational Value Creation Actually Requires

Operational value creation is a phrase that appears in almost every private equity investment thesis and is executed well in relatively few. The gap between aspiration and execution is not primarily a question of intent — most sponsors genuinely want to improve their portfolio companies. It is a question of capability and time.

Genuine operational improvement requires diagnosing the specific constraints that are limiting performance, designing interventions that address those constraints rather than generic best practices, and sustaining the management attention and resource allocation required to see the interventions through. This is harder than it sounds in a portfolio company context, where management bandwidth is limited, the sponsor's involvement is episodic, and the pressure to show short-term results can crowd out the investments that would generate long-term value.

The 100-Day Imperative

The sponsors generating the best operational results are those that treat the first hundred days after acquisition as the most important period of the investment. This is when the organisation is most receptive to change, when the management team is most engaged with the new ownership, and when the diagnostic work that will shape the value creation plan can be done most effectively.

A rigorous hundred-day programme does three things. It establishes a clear baseline of current performance across the dimensions that matter most to value creation. It identifies the two or three operational improvements that will have the largest impact on EBITDA over the investment horizon. And it puts in place the management accountability and reporting cadence that will allow the sponsor to track progress and intervene when the plan is not being executed.

Building for Exit from Day One

The most common mistake in private equity operational improvement is treating it as a separate workstream from exit preparation. Sponsors that begin thinking about what the exit story will need to be only in the final year of the investment consistently find that the operational improvements they have made are not the ones that buyers value most.

The discipline of building for exit from day one means defining the exit thesis at acquisition — what kind of buyer will pay the highest multiple for this business, and what does that buyer need to see — and then designing the operational improvement programme to build the evidence that thesis requires. This alignment between operational work and exit narrative is what separates the sponsors generating top-quartile returns from those generating median ones.

Digitisation Is Not Transformation

The most common failure mode in public sector technology investment is digitising a broken process rather than redesigning it. An agency that moves a paper form online has not improved the citizen experience — it has preserved all of the complexity and friction of the original process while adding the technical overhead of a digital system. The result is a service that is marginally more convenient to access but no faster, no simpler, and no less expensive to operate.

Genuine service modernisation starts with a different question: what outcome does the citizen need, and what is the simplest process that could reliably deliver it? This question frequently reveals that the existing process contains steps that exist for historical reasons — legacy system constraints, organisational boundaries, compliance interpretations that have never been tested — rather than because they are necessary to achieve the outcome.

Designing Around the Citizen Journey

Effective service redesign requires understanding the citizen journey in its entirety, not just the agency's portion of it. A citizen applying for a benefit, renewing a licence, or reporting a problem does not experience the service as a series of agency interactions — they experience it as a single journey that either resolves their need or does not. Agencies that design their services from the citizen's perspective rather than from the agency's organisational structure consistently deliver better outcomes.

This requires a level of cross-agency collaboration that is genuinely difficult in most government structures. Jurisdictional boundaries, data-sharing constraints, and budget accountability that is organised by agency rather than by citizen outcome all create barriers to joined-up service design. The agencies that have overcome these barriers have typically done so through a combination of political mandate and practical demonstration — showing that a joined-up approach delivers better outcomes before asking for the structural changes that would make it permanent.

The Technology Choices That Matter

Public sector technology decisions are frequently dominated by procurement considerations — what can be acquired through existing frameworks, what vendors have the necessary security accreditations, what systems can be integrated with legacy infrastructure. These are legitimate constraints, but they should not drive the service design. The technology choices that matter most are those that determine whether the service can be delivered at the speed and quality citizens expect.

Cloud-native architecture, API-first integration, and modular design are not technology preferences — they are the foundations of a service that can be improved incrementally rather than replaced wholesale every decade. Agencies that have made these architectural choices are finding that they can respond to changing citizen needs and policy requirements far more quickly than those locked into monolithic legacy systems.

Measuring What Matters

Public sector performance measurement has historically focused on inputs and activities — how much was spent, how many transactions were processed, whether the system was available. These metrics are easy to collect but poorly connected to the outcomes that citizens and governments actually care about: whether the service resolved the citizen's need, how long it took, and whether it was delivered at a cost that is sustainable.

Agencies that have shifted to outcome-based measurement are finding that it changes not just how they report performance but how they make decisions. When the metric is resolution rate rather than transaction volume, the incentive is to design services that resolve needs on first contact rather than to process transactions efficiently. That shift in incentive structure is often more powerful than any technology investment.

Six Priorities That Cannot Wait

The agenda facing chief executives in 2026 is not short, but it is not equally weighted. Some decisions can be deferred without material consequence; others are compounding in ways that make delay increasingly costly. The six priorities below are the ones where the cost of inaction is highest and the window for effective action is narrowing.

Priority 1: Commit to an AI Strategy, Not an AI Experiment Portfolio

Most large organisations have accumulated a collection of AI pilots that have not scaled. The problem is rarely the technology — it is the absence of a strategic framework that defines where AI will create the most value, what capabilities the organisation needs to build to capture that value, and how AI deployment will be governed. CEOs who have not yet made that strategic commitment are falling behind peers who have.

Priority 2: Redesign the Workforce Before Automation Does It for You

Automation is eliminating tasks faster than most organisations are redesigning roles. The CEOs managing this transition well are those who have taken a deliberate approach — mapping which tasks will be automated, which roles will be fundamentally changed, and what new capabilities the organisation will need — rather than waiting for the workforce implications of automation to become a crisis.

Priority 3: Build Supply Chain Resilience Into the Capital Plan

Geopolitical risk has permanently changed the calculus of global supply chain design. The organisations that treated the disruptions of recent years as temporary anomalies and returned to pre-disruption network designs are the most exposed. Building resilience requires capital — for alternative sourcing, for strategic inventory, for nearshoring — and that capital needs to be in the plan, not treated as an optional response to the next crisis.

Priority 4: Treat Sustainability as a Strategic Input, Not a Reporting Obligation

The organisations that are generating competitive advantage from sustainability are those that have integrated it into strategic planning — using it to identify cost reduction opportunities, to anticipate regulatory requirements before they become binding, and to differentiate their value proposition with customers who are making purchasing decisions on sustainability grounds. The organisations treating it as a reporting obligation are incurring the cost without capturing the benefit.

Priority 5: Invest in the Management Layer

The quality of the management layer — the leaders below the C-suite who translate strategy into execution — is the most reliable predictor of whether a strategic agenda gets implemented. CEOs who have invested in developing this layer, through deliberate succession planning, structured development programmes, and honest performance management, consistently outperform those who have not.

Priority 6: Clarify the Capital Allocation Framework

In a higher-rate environment, the cost of capital is real and the opportunity cost of misallocation is significant. CEOs who have not revisited their capital allocation framework since the low-rate era are likely making investment decisions against hurdle rates that no longer reflect the true cost of capital. Clarifying the framework — and being willing to exit businesses and investments that do not meet it — is one of the highest-value decisions available to a chief executive in the current environment.

Five Forces Converging on the Global Network

Supply chain strategy is being reshaped by five converging forces that individually would each warrant attention. Together, they are driving a fundamental rethink of how global networks are designed, operated, and governed. The organisations that are managing this transition well are those that have recognised the forces as interconnected and are responding with an integrated strategy rather than five separate workstreams.

Trend 1: Nearshoring and Regional Network Design

The economics of global sourcing have shifted. Rising labour costs in traditional low-cost manufacturing locations, combined with the risk premium that long supply chains now carry, have made nearshoring attractive for a wider range of products than was the case a decade ago. The organisations moving fastest on nearshoring are those that have done the rigorous cost analysis — accounting for total landed cost, not just unit cost — and have identified the specific product categories where regional production is now competitive.

Trend 2: Digital Twins for Network Optimisation

Digital twin technology has matured to the point where it is genuinely useful for supply chain decision-making, not just simulation. Organisations that have built digital representations of their supply networks are using them to model the impact of disruptions before they occur, to optimise inventory positioning across the network, and to evaluate the trade-offs between cost and resilience in network design decisions. The value is not in the technology itself but in the quality of the decisions it enables.

Trend 3: AI-Driven Demand Sensing

Traditional demand forecasting relies on historical sales data and is inherently backward-looking. AI-driven demand sensing incorporates a broader range of signals — point-of-sale data, search trends, social signals, weather patterns, economic indicators — to generate forecasts that are more accurate and more responsive to emerging demand shifts. The organisations that have deployed demand sensing effectively have reduced both stockouts and excess inventory, improving service levels and working capital simultaneously.

Trend 4: Supplier Relationship Transformation

The transactional supplier relationship model — lowest price wins, switch suppliers when a better price is available — is being replaced by a more collaborative model in which key suppliers are treated as strategic partners. This shift is driven partly by the recognition that resilience requires supplier investment and cooperation, and partly by the realisation that the best suppliers have choices about which customers they prioritise in a constrained environment.

Trend 5: Sustainability as a Supply Chain Constraint

Sustainability requirements are moving from voluntary commitments to binding constraints in supply chain design. Regulatory requirements for supply chain emissions disclosure, customer requirements for verified sustainable sourcing, and the financial implications of carbon pricing are all creating pressure to redesign networks around sustainability criteria as well as cost and service. The organisations that have integrated sustainability into their supply chain strategy are finding that it drives innovation in network design rather than simply adding cost.

From Compliance to Capability

The framing of AI governance as a compliance requirement has been damaging to the organisations that have adopted it. Compliance framing creates a minimisation incentive — do the minimum required to satisfy the regulator, and treat anything beyond that as unnecessary cost. Applied to AI governance, this produces frameworks that are technically adequate and practically useless: policies that exist but are not followed, review processes that approve everything, and accountability structures that assign responsibility without authority.

The organisations that are generating competitive advantage from AI governance have adopted a different framing. They treat governance as the capability that allows them to deploy AI at scale with confidence — confidence that the models are performing as intended, that the risks are understood and managed, and that the organisation can explain its AI decisions to any audience that asks. That confidence is itself a competitive asset.

What Effective Governance Enables

Effective AI governance enables three things that compliance-oriented governance does not. First, it enables faster deployment. Organisations with clear governance frameworks can move from use-case identification to production deployment more quickly because the review process is well-defined and proportionate — high-risk use cases get rigorous review, low-risk ones get expedited approval. Second, it enables broader deployment. Organisations that have built trust with their boards and regulators through demonstrated governance discipline have more latitude to deploy AI in sensitive domains.

Third, it enables learning. Organisations with effective governance frameworks capture what they learn from each deployment — what worked, what did not, what risks materialised — and use that learning to improve subsequent deployments. This creates a compounding advantage: each deployment makes the next one better, and the organisation's AI capability improves faster than that of competitors who are treating each deployment as a one-off project.

The Board Dimension

AI governance has become a board-level issue in most large organisations, and the boards that are engaging with it most effectively are those that have moved beyond asking whether AI is being used responsibly to asking whether the organisation's AI governance capability is a source of competitive advantage. This is a more demanding question, and it requires a more sophisticated answer than most management teams are currently prepared to give.

The management teams that are prepared to give that answer have invested in board education — helping directors understand what AI governance requires, what good looks like, and what questions to ask. They have also built the reporting infrastructure to give the board visibility into AI deployment and performance across the organisation, not just in the high-profile use cases that make it into board presentations.

Building the Governance Advantage

Building AI governance as a competitive advantage requires treating it as an organisational capability rather than a policy document. It requires investment in the people, processes, and tools that make governance operational — model risk management teams with genuine authority, monitoring infrastructure that detects model drift and bias in production, and escalation processes that work in practice rather than on paper.

The organisations that have made this investment are finding that it pays dividends beyond AI. The discipline of defining accountability, monitoring performance, and learning from failure that effective AI governance requires is the same discipline that drives performance improvement across the organisation. AI governance, done well, is not a cost of doing AI — it is a capability that makes the whole organisation better.

The Automation Productivity Paradox

Organisations have been investing in automation for decades, and the aggregate productivity gains have been disappointing relative to the investment. The paradox is not hard to explain: automation that eliminates individual tasks without changing the workflow those tasks are embedded in does not improve productivity — it creates gaps in the workflow that humans fill with workarounds, manual reconciliation, and coordination overhead that was not there before.

The organisations that are generating real productivity gains from automation are those that treat it as an opportunity to redesign work rather than to automate existing work. The distinction sounds simple but requires a fundamentally different approach to implementation — one that starts with the desired workflow and works backward to the technology, rather than starting with the technology and asking what it can automate.

Role Redesign as the Critical Enabler

The most important and most neglected element of automation-driven productivity improvement is role redesign. When automation eliminates tasks from a role, the default response is to add other tasks to fill the time — tasks that are available rather than tasks that are valuable. The result is a role that is busier but not more productive, and a workforce that is more stressed but not more capable.

Effective role redesign starts with a clear view of what the role should accomplish — what decisions it should make, what relationships it should manage, what value it should create — and then designs the task set around those outcomes. This often means eliminating tasks that are currently performed but do not contribute to the role's core purpose, as well as adding tasks that the role should be performing but currently is not because there is no time.

The Measurement Problem

Productivity measurement in knowledge work is genuinely difficult, and most organisations are measuring the wrong things. Activity metrics — emails sent, meetings attended, tasks completed — are easy to collect but poorly correlated with the outcomes that matter. The organisations that are managing productivity most effectively have invested in defining outcome metrics for each role and measuring those rather than activity proxies.

This requires a level of clarity about what each role is supposed to accomplish that many organisations lack. Job descriptions that describe activities rather than outcomes, performance management systems that reward effort rather than results, and management cultures that equate presence with productivity all create barriers to effective productivity measurement. Addressing these barriers is harder than deploying automation, but it is the prerequisite for capturing the productivity gains that automation makes possible.

The Human Capability Investment

The productivity gains available from automation are ultimately bounded by the capability of the humans working alongside the automated systems. An organisation that has automated its routine processes but has not invested in developing the judgment, creativity, and relationship skills that the remaining human roles require will find that its productivity ceiling is lower than it expected.

The organisations generating the highest productivity returns from automation are those that have treated it as an opportunity to upgrade the human capability in their workforce — investing in the skills that automation cannot replicate and creating roles that are genuinely more demanding and more rewarding than the ones they replaced. That investment in human capability is what converts automation from a cost-reduction tool into a genuine productivity multiplier.

Why the Old Playbook Is Insufficient

The private equity value creation playbook that dominated the low-rate era — buy at a reasonable multiple, add leverage, grow revenue, and exit at a higher multiple — is no longer sufficient. The multiple expansion that made the financial engineering component of that playbook so powerful has largely reversed. The leverage that amplified returns is now more expensive. And the exit environment rewards businesses that can demonstrate genuine operational improvement rather than financial structuring.

The sponsors that are generating top-quartile returns in the current environment have updated their playbook. They are still disciplined acquirers, but they are generating returns primarily through operational improvement rather than financial engineering. And they are doing it systematically — with structured programmes, dedicated resources, and clear accountability — rather than opportunistically.

The Structured Improvement Programme

A structured operational improvement programme is not a list of initiatives — it is a management system. It starts with a rigorous diagnostic that identifies the specific operational constraints limiting EBITDA, not the generic improvement opportunities that appear in every investment thesis. It translates those constraints into a prioritised set of initiatives with clear owners, milestones, and financial targets. And it establishes a governance cadence that gives the sponsor visibility into progress and early warning when initiatives are off track.

The diagnostic phase is where most programmes fail. The temptation is to move quickly from acquisition to implementation — to start executing the value creation plan that was developed during due diligence rather than testing its assumptions against the reality of the business. The sponsors generating the best results are those that invest the first sixty to ninety days in rigorous diagnosis, even when it means delaying the start of implementation.

Revenue Growth as an Operational Challenge

Revenue growth in private equity portfolio companies is frequently treated as a commercial challenge — a question of sales force effectiveness, pricing strategy, and market positioning. These are important, but the most durable revenue growth comes from operational improvements that make the business genuinely better for customers: faster delivery, higher quality, more consistent service, better product availability.

Sponsors that have recognised this are investing in operational improvements that directly improve the customer experience, not just the cost structure. They are finding that these investments generate revenue growth as well as margin improvement, and that the combination produces the kind of EBITDA trajectory that commands premium exit multiples.

The Talent Imperative

Operational improvement programmes succeed or fail based on the quality of the management team executing them. The sponsors generating the best results are those that make management assessment and development a priority from day one — identifying the leaders who have the capability to drive the improvement programme, filling gaps quickly when they find them, and investing in the development of the leaders who will carry the business through to exit.

This requires a level of engagement with talent that goes beyond the traditional private equity model of hiring a strong CEO and delegating. The sponsors that are winning are those that treat talent as a portfolio-level capability — building the assessment tools, development programmes, and executive networks that allow them to upgrade management quality systematically across their portfolio.

Beyond Process Automation

The first wave of operational automation was about eliminating manual steps in well-defined processes. Robotic process automation, workflow tools, and enterprise resource planning systems all delivered value by replacing human effort in tasks that were repetitive, rule-based, and high-volume. That wave has largely run its course in most large industrial organisations — the obvious automation opportunities have been captured, and the remaining manual work is either genuinely complex or embedded in processes that are too fragmented to automate cleanly.

The second wave is different in kind. AI-enabled process orchestration does not just automate individual tasks — it coordinates across tasks, adapts to changing conditions, and makes decisions that previously required human judgment. The result is not just a faster version of the existing process but a fundamentally different operating model in which the system manages the workflow and humans focus on the exceptions and the decisions that genuinely require their judgment.

Where the Value Is Being Created

The industrial firms that have moved intelligent operations from pilot to production are generating value in three areas. First, cycle time compression: AI-orchestrated processes can respond to changes in demand, supply, or operating conditions in real time, eliminating the delays that accumulate in human-mediated handoffs. Second, quality improvement: AI systems that monitor process parameters continuously can detect quality deviations earlier and more reliably than periodic human inspection, reducing the cost of defects and rework.

Third, and most significantly, capacity optimisation: AI systems that understand the full complexity of a production or logistics network can identify utilisation opportunities that are invisible to human planners working with simplified models. The organisations that have deployed AI for capacity optimisation are finding that they can serve more volume with the same assets — a direct improvement in return on capital.

The Integration Challenge

The primary obstacle to scaling intelligent operations is not the AI technology — it is the integration challenge. AI systems that orchestrate processes need access to real-time data from across the operation: production systems, logistics platforms, quality management systems, and supplier data feeds. In most industrial organisations, these systems were not designed to share data in real time, and integrating them requires significant engineering effort.

The organisations that have solved this problem have typically done so by building a data integration layer — an operational data platform that aggregates and normalises data from across the operation — before deploying AI applications on top of it. This sequencing is important: AI applications built on fragmented, inconsistent data deliver inconsistent results, and the resulting loss of confidence in the technology can set back the broader intelligent operations programme by years.

The Human Role in Intelligent Operations

The transition to intelligent operations raises legitimate questions about the human role in the operating model. The honest answer is that the roles that are most at risk are those that consist primarily of information relay and routine decision-making — roles that exist because humans are needed to move information between systems and make decisions that the systems cannot make. As AI systems become capable of doing both, those roles will change significantly.

The organisations managing this transition well are those that are investing in redefining the human role around the capabilities that AI cannot replicate: judgment in novel situations, relationship management with customers and suppliers, creative problem-solving when the system encounters conditions it was not designed for. This is not a comfortable transition for everyone, but it is the honest one, and the organisations that are having that conversation with their workforce are managing it better than those that are not.

The Limits of the First Wave

Enterprise automation's first wave delivered real but bounded value. Robotic process automation reduced the cost of high-volume, rule-based tasks. Workflow tools eliminated manual handoffs in well-defined processes. These investments were justified and have largely paid off. But they have also revealed the limits of task-level automation: when you automate individual tasks without redesigning the process they are part of, you create a faster version of a broken process rather than a better one.

The organisations that extracted the most value from first-wave automation are now confronting a ceiling. The remaining manual work is either genuinely complex, embedded in processes that span organisational boundaries, or dependent on judgment that current automation tools cannot replicate. Moving beyond that ceiling requires a different approach — one that starts with process redesign rather than task automation.

End-to-End Process Redesign

Second-wave automation is characterised by end-to-end process redesign — taking a process from its trigger to its outcome and asking what the ideal version of that process looks like if technology constraints are removed. This question frequently reveals that the existing process contains steps that exist only because of legacy system limitations, organisational boundaries that made sense in a different operating model, or compliance interpretations that have never been tested.

The processes that are most amenable to this kind of redesign are those that currently involve multiple handoffs between systems and people, where each handoff introduces delay and the possibility of error. Order-to-cash, procure-to-pay, hire-to-retire — the classic enterprise processes that have been partially automated but never fundamentally redesigned — are the primary targets for second-wave automation.

The Structural Implications

End-to-end process redesign has structural implications that task-level automation does not. When a process is redesigned around automation rather than around the organisational units that currently own its components, the question of who owns the redesigned process does not have an obvious answer. The finance team owns the payment step, the procurement team owns the purchase order step, and the operations team owns the receiving step — but the end-to-end process does not belong to any of them.

The organisations that are managing this structural challenge most effectively are those that have created process ownership roles that cut across functional boundaries — leaders who are accountable for the end-to-end performance of a process rather than for the performance of a functional unit. This is a significant organisational change, and it requires executive sponsorship and a willingness to challenge the functional accountability model that most large organisations are built around.

What Comes Next

The third wave of enterprise automation — agentic AI systems that can manage complex, multi-step processes with minimal human intervention — is already visible in early deployments. These systems are not yet reliable enough for most enterprise applications, but they are improving rapidly, and the organisations that are building the process clarity and data infrastructure that second-wave automation requires are also building the foundation for third-wave deployment.

The strategic implication is that automation investment should be evaluated not just on its immediate return but on the capabilities it builds for future waves. An organisation that invests in end-to-end process redesign today is not just improving current performance — it is building the process clarity, data quality, and organisational capability that will allow it to deploy agentic AI effectively when the technology is ready.

The Problem with Point Forecasts

Most strategic planning processes are built around a point forecast — a single view of how the economic environment will evolve over the planning horizon. This approach has the virtue of simplicity: it produces a clear set of targets and a coherent resource allocation. But it has a fundamental flaw: the forecast is almost certainly wrong, and the planning process provides no guidance on what to do when it is.

The organisations that navigate economic uncertainty most effectively are those that have replaced point forecasts with scenario frameworks — structured views of how the environment might evolve across a range of plausible futures. The value of scenario planning is not that it predicts the future accurately; it is that it forces the organisation to think through the implications of different futures before they arrive, and to identify the decisions that are robust across scenarios versus those that are highly sensitive to which scenario materialises.

Building Scenarios That Drive Decisions

Scenario planning is widely practised and widely misused. The most common failure mode is producing scenarios that are too similar to each other to generate genuinely different strategic implications — scenarios that differ in degree rather than in kind, and that therefore converge on the same strategic recommendations. Useful scenarios are those that are genuinely distinct, that reflect different underlying assumptions about the forces that matter most, and that lead to different conclusions about what the organisation should do.

Building scenarios that drive decisions requires starting with the uncertainties that matter most — the variables whose resolution will have the largest impact on the organisation's strategic position — and constructing scenarios around different resolutions of those uncertainties. This is harder than it sounds, because the variables that matter most are often the ones that are hardest to forecast, and the temptation is to build scenarios around variables that are easier to model even if they are less consequential.

Capital Flexibility as Strategic Infrastructure

Scenario planning is only valuable if it is connected to capital allocation. An organisation that has developed a sophisticated scenario framework but has committed all of its capital to a single strategic direction has not actually prepared for uncertainty — it has just described it more articulately. Capital flexibility — the ability to redirect investment quickly when conditions change — is the operational expression of scenario preparedness.

Building capital flexibility requires discipline in the current period: maintaining balance sheet capacity, avoiding commitments that cannot be unwound, and being willing to exit investments that are not performing rather than holding them in the hope that conditions will improve. These are difficult decisions in an environment where growth is being rewarded, but they are the decisions that create the optionality to move decisively when the environment shifts.

Moving Decisively When Conditions Shift

The organisations that generate the most value from economic uncertainty are not those that predict it most accurately — they are those that move most decisively when conditions shift. This requires two things: the strategic clarity to know what to do when a scenario materialises, and the organisational agility to act on that clarity quickly.

Strategic clarity comes from the scenario planning work — having thought through in advance what the organisation would do in each scenario, so that when a scenario begins to materialise the response is not improvised. Organisational agility comes from the capital flexibility and the decision-making processes that allow the organisation to act on its strategic clarity without being slowed by the inertia that affects most large organisations in periods of change.

Why Cyber Risk Is a Board Issue

Cyber risk has crossed the threshold from a technical concern to a governance concern. The consequences of a significant cyber incident — operational disruption, regulatory penalty, reputational damage, and in some cases existential threat to the business — are squarely within the range of risks that boards are responsible for overseeing. The question is no longer whether boards should engage with cyber risk but how to engage with it effectively.

The challenge is that most boards are not well equipped to engage with cyber risk. Directors who are comfortable evaluating financial risk, strategic risk, and operational risk often find cyber risk opaque — the technical language is unfamiliar, the threat landscape changes rapidly, and the connection between technical controls and business outcomes is not always clear. The result is board engagement that is superficial: asking whether the organisation has been breached rather than whether the organisation's cyber risk posture is appropriate for its strategic context.

Translating Cyber Risk for the Board

The management teams that are engaging their boards most effectively on cyber risk have made a deliberate investment in translation — converting technical cyber risk information into business risk language that directors can evaluate against the organisation's risk appetite. This means expressing cyber risk in terms of business impact rather than technical severity, connecting specific vulnerabilities to specific business consequences, and presenting the organisation's cyber risk posture in the context of its strategic priorities.

It also means being honest about uncertainty. Cyber risk is genuinely difficult to quantify, and management teams that present false precision — specific probability estimates for specific attack scenarios — are not helping their boards make better decisions. The more useful approach is to present a range of scenarios, describe the controls in place to reduce the likelihood and impact of each, and be clear about the residual risk that the organisation is accepting.

Cyber Risk as a Strategic Input

The organisations managing cyber risk most effectively are those that have integrated it into strategic decision-making rather than treating it as a separate risk management function. When the organisation is evaluating a major technology investment, a merger or acquisition, or a new market entry, cyber risk should be part of the analysis — not as a veto, but as a factor that shapes the design of the initiative and the investment required to execute it safely.

This integration requires cyber risk professionals who can engage with strategic questions rather than just technical ones, and business leaders who understand enough about cyber risk to ask the right questions. Building that mutual capability is a longer-term investment, but it is the foundation of an organisation that manages cyber risk as a strategic input rather than a technical afterthought.

The Resilience Imperative

The most important shift in cyber risk thinking over the past several years has been from prevention to resilience. The assumption that a sufficiently well-defended organisation will not be breached has been replaced by the recognition that breaches are inevitable, and that the organisations that manage cyber risk best are those that can detect, contain, and recover from incidents quickly rather than those that have the most sophisticated perimeter defences.

Resilience requires investment in detection and response capabilities that many organisations have underweighted relative to prevention. It also requires regular testing — not just of technical controls but of the organisational response to a significant incident. The organisations that have invested in cyber resilience are finding that the discipline of preparing for incidents they hope will never happen is making them better at managing the ones that do.

Resilience Is Not Robustness

Organisational resilience is frequently confused with organisational robustness. A robust organisation is one that is built to withstand stress without changing — strong balance sheet, diversified revenue, redundant operations. These are valuable properties, but they are not the same as resilience. A resilient organisation is one that can absorb a disruption, adapt its operating model in response, and emerge from the disruption in a stronger competitive position than it entered.

The distinction matters because the investments required to build robustness and resilience are different. Robustness is built through financial and operational redundancy. Resilience is built through decision-making capability, cultural adaptability, and the organisational learning systems that allow an organisation to update its understanding of the environment and respond accordingly.

Decision-Making Speed Under Pressure

The organisations that respond most effectively to unexpected disruptions are those that can make good decisions quickly under conditions of uncertainty and incomplete information. This capability is not primarily a function of having the right decision-making processes — it is a function of having leaders who are comfortable making decisions without complete information, and an organisational culture that supports rapid decision-making rather than demanding consensus before acting.

Building this capability requires deliberate practice. Organisations that run regular scenario exercises — simulating disruptions and practising the decision-making process under pressure — develop the muscle memory that allows them to respond effectively when a real disruption occurs. Organisations that have never practised disruption response discover in the moment that their decision-making processes, which work adequately in normal conditions, are too slow and too consensus-dependent to be effective in a crisis.

Cultural Adaptability

Culture is the most powerful determinant of organisational resilience and the hardest to change. Cultures that value stability, predictability, and adherence to established processes are well suited to stable environments but poorly suited to disrupted ones. Cultures that value learning, experimentation, and adaptation are better suited to disrupted environments but may sacrifice some efficiency in stable ones.

The organisations that have built genuinely resilient cultures are those that have found a way to maintain operational discipline in normal conditions while preserving the cultural flexibility to adapt when conditions change. This is a difficult balance to strike, and it requires leaders who model the behaviour they want — who demonstrate that it is acceptable to change course when the evidence demands it, and who treat the lessons from disruptions as valuable inputs to organisational learning rather than as failures to be explained away.

Learning as a Resilience Capability

The most resilient organisations are those that learn faster than their environment changes. This requires systematic investment in organisational learning — the processes and practices that allow the organisation to capture what it learns from experience, distribute that learning across the organisation, and update its operating model in response.

Most organisations have learning processes that work adequately in stable conditions but break down under stress. After-action reviews that are conducted months after the event, lessons-learned databases that no one reads, and knowledge management systems that capture information but do not change behaviour — these are the learning infrastructure of an organisation that is not serious about resilience. Building genuine learning capability requires treating it as an operational priority rather than an administrative function.

The Hurdle Rate Problem

The decade of near-zero interest rates that preceded 2022 created a generation of capital allocation frameworks built around a cost of capital that no longer exists. Hurdle rates that were set in a low-rate environment systematically understate the true cost of capital in the current environment, and organisations that have not updated them are approving investments that would not pass a rigorous cost-of-capital test.

The consequences are not immediately visible — a project approved at a hurdle rate that is too low will not obviously underperform in its early years. But the cumulative effect of systematically misallocating capital over several years is a portfolio of investments that is generating returns below the cost of capital, eroding shareholder value in ways that are difficult to attribute to any single decision.

Updating the Framework

Updating a capital allocation framework is straightforward in principle and politically difficult in practice. The technical work — recalculating the weighted average cost of capital, updating hurdle rates, and re-evaluating the existing investment portfolio against the new rates — is not complicated. The difficult part is the organisational conversation that follows: some investments that were approved under the old framework will not pass the new one, and some business units that were generating acceptable returns in a low-rate environment will not be generating acceptable returns in the current one.

The organisations that have managed this conversation most effectively are those that have framed the framework update as a response to a changed environment rather than as a judgment on past decisions. The investments that were approved under the old framework were not necessarily wrong — they were made with the information available at the time. The question is what to do with them now, given that the environment has changed.

Portfolio Rationalisation

A rigorous capital allocation framework in a higher-rate environment will typically identify a set of investments and business units that are not generating returns above the cost of capital. The appropriate response is portfolio rationalisation — exiting those investments and redeploying the capital to higher-return opportunities. This is a discipline that most organisations find difficult, because the businesses and investments that are underperforming are typically ones that have internal advocates and historical significance.

The organisations that have executed portfolio rationalisation most effectively are those that have made it a systematic process rather than a crisis response. Regular portfolio reviews that evaluate each business and investment against the current cost of capital, with clear criteria for what triggers an exit decision, are more effective than ad hoc reviews triggered by financial distress.

The Opportunity in Discipline

The higher-rate environment is not only a constraint — it is also an opportunity for organisations that have the discipline to maintain rigorous capital allocation standards. In an environment where many organisations are misallocating capital, the organisations that are allocating it well will generate superior returns and will have the financial capacity to invest in the opportunities that their less disciplined competitors cannot afford to pursue.

The organisations that will emerge from the current environment in the strongest competitive position are those that have used the higher-rate period to build the capital allocation discipline that the low-rate era allowed them to avoid. That discipline — clear hurdle rates, rigorous portfolio reviews, and the willingness to exit underperforming investments — is not just a response to the current environment. It is the foundation of sustained value creation in any environment.

The Real Reason AI Strategies Fail

When AI initiatives fail to deliver their promised value, the post-mortem almost always identifies the same root cause: the data was not good enough. Not the model, not the algorithm, not the compute infrastructure — the data. The training data was incomplete, inconsistent, or biased. The production data was structured differently from the training data. The data pipeline was unreliable, introducing latency or errors that degraded model performance. The labels were wrong.

This is not a new finding. The data quality problem in AI has been documented extensively, and most organisations are aware of it in the abstract. The problem is that awareness has not translated into action. Data infrastructure investment is unglamorous, slow to show results, and difficult to connect to the business outcomes that justify capital allocation. AI model investment is exciting, fast-moving, and easy to demonstrate in a proof of concept. The result is organisations that have invested heavily in AI models and lightly in the data foundations those models require.

What a Data Foundation Actually Requires

A data foundation that can support enterprise AI is not a single system — it is a set of capabilities that work together. Data integration: the ability to bring data from disparate sources into a common environment where it can be used together. Data quality management: the processes and tools to identify, measure, and remediate data quality issues before they propagate into AI models. Data governance: the policies and accountabilities that determine who can access what data, how it can be used, and how long it can be retained.

Most organisations have some version of each of these capabilities, but they are typically fragmented — different tools for different domains, governance policies that are inconsistently applied, and data quality management that is reactive rather than systematic. Building a data foundation that can support enterprise AI requires consolidating and strengthening these capabilities, which is a multi-year programme rather than a project.

The Sequencing Question

The most common question we hear from organisations that have recognised the data foundation problem is how to sequence the investment: should they fix the data foundation before deploying AI, or deploy AI in parallel with fixing the data foundation? The honest answer is that it depends on the use case and the current state of the data.

For use cases where data quality is critical — credit decisions, medical diagnosis, fraud detection — the data foundation needs to be adequate before deployment. For use cases where the consequences of model errors are lower and the learning value of deployment is high — content recommendation, internal search, document summarisation — deploying in parallel with data improvement can be justified. The key is being honest about which category each use case falls into, rather than defaulting to the more exciting option.

Making the Investment Case

The investment case for data foundation work is harder to make than the investment case for AI model development, because the returns are indirect and delayed. Data infrastructure does not generate value by itself — it generates value by enabling AI applications that would not otherwise be possible, or by improving the performance of AI applications that are already deployed.

The organisations that have made this investment case successfully are those that have connected specific data foundation investments to specific AI use cases with quantified value, and have presented the data investment as the prerequisite for the AI value rather than as a separate initiative. This framing — data investment as AI enabler rather than data investment as IT infrastructure — is more compelling to business leaders and more likely to secure the funding that the work requires.

The Grid as the Binding Constraint

The energy transition has a bottleneck, and it is not renewable generation capacity or battery storage technology. It is the grid. Transmission infrastructure that was designed for a centralised generation model cannot efficiently move power from distributed renewable sources to load centres. Distribution infrastructure that was built for one-way power flow cannot manage the two-way flows that rooftop solar, electric vehicles, and distributed storage require. And the interconnection queues that determine when new generation can connect to the grid are measured in years, not months.

This is not a new problem, but it is an increasingly urgent one. The pace of renewable generation development has outrun the pace of grid investment, creating a growing mismatch between generation capacity and the infrastructure needed to deliver it. Utilities that are not treating grid modernisation as a strategic priority are not just falling behind on their own transition plans — they are becoming a constraint on the broader energy transition.

Transmission: The Long-Lead Investment

Transmission investment is the most consequential and most difficult grid investment to make. The lead times are long — a major transmission project can take a decade from planning to operation — the permitting process is complex, and the cost allocation questions are genuinely difficult. Who pays for a transmission line that benefits multiple states and multiple utilities is not a question with an obvious answer, and the absence of a clear answer has been a significant barrier to investment.

The utilities that are making progress on transmission investment are those that have invested in the regulatory and stakeholder engagement work that is the prerequisite for project approval, not just the engineering work. Transmission projects that fail do so more often because of permitting and stakeholder opposition than because of technical or financial challenges. Building the relationships and the regulatory strategy that allow a project to navigate those obstacles is as important as building the project itself.

Distribution: The Intelligent Grid

Distribution grid modernisation is a different kind of challenge from transmission. The investment is more granular, the technology is more mature, and the business case is more straightforward. Advanced metering infrastructure, distribution automation, and grid management software are all proven technologies with clear operational benefits. The challenge is the scale of the investment required and the pace at which it needs to happen.

The utilities that are executing distribution modernisation most effectively are those that have developed a clear prioritisation framework — identifying the parts of the distribution network where modernisation will deliver the highest value, whether measured in reliability improvement, capacity for distributed energy resources, or support for electric vehicle charging. Trying to modernise everything simultaneously is neither financially feasible nor operationally manageable; a disciplined prioritisation approach is essential.

The Strategic Opportunity

Grid modernisation is not just a cost — it is a platform for new business models. Utilities that have built intelligent, flexible grid infrastructure are finding that it enables services that were not previously possible: demand response programmes that pay customers to reduce consumption at peak times, vehicle-to-grid services that use electric vehicle batteries as distributed storage, and real-time pricing that allows customers to optimise their energy use.

These new services are not yet large revenue contributors for most utilities, but they are growing, and the utilities that have built the grid infrastructure to support them are better positioned to capture that growth than those that have not. The strategic opportunity in grid modernisation is not just to reduce the cost of delivering existing services — it is to build the platform for the energy services of the next decade.

The Compliance Trap

Risk management in the public sector has been shaped by a compliance imperative that has produced frameworks that are technically adequate and practically limited. Agencies have risk registers, risk committees, and risk reporting processes. They satisfy audit requirements and produce the documentation that oversight bodies expect. But the risk management function rarely influences the decisions that matter most — where to invest, how to design services, which programmes to prioritise — because it is positioned as a compliance function rather than a strategic one.

The consequence is that risk management adds cost without adding value. It consumes management time in reporting processes that do not change decisions, and it creates a false sense of security — the appearance of risk management without the substance. The agencies that have broken out of this trap are those that have repositioned risk management as a tool for better decision-making rather than a compliance obligation.

Risk Management as a Decision Tool

Effective risk management in the public sector starts with a clear understanding of what decisions it is meant to inform. The most valuable application is investment decision-making: helping agencies understand the risk profile of different investment options, the conditions under which each option is most likely to succeed, and the mitigation strategies that can improve the probability of success.

This requires risk analysis that is specific to the decision at hand rather than generic. A risk register that lists the same categories of risk for every programme — financial risk, operational risk, reputational risk — does not help a decision-maker choose between two investment options with different risk profiles. Useful risk analysis identifies the specific risks that are material to the specific decision, quantifies them to the extent possible, and connects them to the decision criteria that the agency is using.

Building Public Trust Through Transparency

Public agencies operate in an environment of heightened scrutiny, and the way they manage and communicate risk has a direct impact on public trust. Agencies that are transparent about the risks they are managing — that acknowledge uncertainty, explain their mitigation strategies, and report honestly when risks materialise — build more trust than those that project false confidence.

This is counterintuitive for many public sector leaders, who have been trained to project certainty and avoid admitting uncertainty. But the public's tolerance for honest uncertainty is higher than most leaders assume, and the reputational damage from a risk that materialises without warning is far greater than the reputational cost of acknowledging that a programme carries risk. Transparency about risk is not a weakness — it is a foundation of credibility.

The Capability Investment

Repositioning risk management as a strategic function requires investment in capabilities that most public sector risk functions do not currently have. Quantitative risk analysis, scenario modelling, and the ability to connect risk analysis to financial and operational decision-making are skills that are in short supply in most government risk teams.

Building these capabilities requires a combination of recruitment, training, and organisational design. Risk professionals who can engage with strategic questions need to be positioned close to decision-makers, not in a separate compliance function. And the decision-making processes that risk management is meant to inform need to be designed to create space for risk input rather than treating it as a box to be checked after the decision has already been made.

What Talent Wants Now

The talent market has changed in ways that most organisations have not fully absorbed. The candidates who are most in demand — those with the technical skills, judgment, and adaptability that the automated economy requires — are not primarily motivated by compensation. They are motivated by the quality of the work, the pace of their own development, and the clarity of the path from where they are to where they want to be.

Organisations that are winning the talent competition in this environment are those that have invested in making the work genuinely better — more interesting, more consequential, more developmental — rather than simply offering higher salaries. This is not an argument against competitive compensation; it is an argument that compensation alone is insufficient to attract and retain the talent that the automated economy requires.

Designing for Development

The most powerful talent retention tool available to most organisations is a credible development path — a clear and honest account of how the organisation will invest in an individual's growth, what skills they will develop, and what roles they will be prepared for. This is more powerful than compensation because it addresses the question that high-potential talent is actually asking: not just what will I earn here, but who will I become?

Building credible development paths requires organisations to be honest about what they can offer. An organisation that promises development but provides no structured learning, no stretch assignments, and no feedback is making a promise it cannot keep, and the talent it attracts on the basis of that promise will leave when the reality becomes clear. The organisations that retain the best talent are those that have built development infrastructure — mentoring, structured learning, deliberate assignment design — that delivers on the promise.

The Automation Anxiety Problem

Automation anxiety — the fear that one's role will be automated away — is a genuine barrier to talent engagement in many organisations. Employees who believe that their jobs are at risk from automation are less likely to invest in developing the skills that the organisation needs, less likely to engage with AI tools that could improve their productivity, and more likely to leave for organisations where they feel more secure.

The organisations that are managing automation anxiety most effectively are those that are having honest conversations about which roles will change and how, rather than offering reassurances that are not credible. Employees who understand that their role will change, and who are being supported in developing the skills that the changed role will require, are more engaged and more productive than those who are being told that nothing will change.

Building for the Long Term

The organisations that will have the talent advantage in the automated economy are those that are making long-term investments in human capability rather than treating talent as a variable cost to be optimised in the short term. This means investing in learning and development even when the immediate return is not obvious, maintaining employment stability through economic cycles rather than using layoffs as the primary cost management tool, and building the organisational culture that attracts and retains people who have choices.

These investments are harder to justify in a quarterly earnings framework than in a multi-year strategic framework. But the organisations that have made them are finding that the talent advantage they have built is one of the most durable competitive advantages available — harder to replicate than technology, harder to buy than market position, and more valuable in an environment where the quality of human judgment is the primary differentiator.

From Simulation to Operations

Digital twin technology entered manufacturing as a simulation tool — a way to model production processes, test changes in a virtual environment, and reduce the cost of physical prototyping. That application has delivered real value, but it has also created a perception of digital twins as a design-phase tool rather than an operational one. The manufacturers that are generating the most value from digital twins have moved beyond simulation to live operational use — using real-time digital representations of their production systems to make better decisions in the moment.

The shift from simulation to live operations requires a different kind of digital twin. A simulation model can be updated periodically and does not need to reflect the current state of the physical system with precision. A live operational twin needs to be continuously updated with real-time data from sensors, production systems, and quality management tools, and it needs to be integrated with the decision-making processes that operators and managers use every day.

Throughput Optimisation in Practice

The most immediate value that live digital twins deliver in manufacturing is throughput optimisation. A digital twin that reflects the current state of a production line — machine status, work-in-progress inventory, quality hold rates, maintenance schedules — can identify bottlenecks and optimisation opportunities that are invisible to operators working with lagged data and simplified mental models of the system.

The manufacturers that have deployed live digital twins for throughput optimisation are finding that the gains come not from dramatic interventions but from the accumulation of small improvements — adjustments to scheduling, sequencing, and buffer management that individually seem minor but collectively add up to significant throughput improvement. The digital twin makes these adjustments visible and actionable in a way that human intuition and periodic reporting cannot.

Quality Management at Scale

Quality management is the second major application area where live digital twins are delivering value. A digital twin that integrates process parameters, environmental conditions, and quality inspection results can identify the combinations of conditions that are associated with quality failures before those failures occur — allowing operators to adjust process parameters proactively rather than responding to defects after the fact.

This predictive quality capability is particularly valuable in complex manufacturing processes where the relationship between process parameters and quality outcomes is not fully understood. The digital twin learns from the data it accumulates, improving its predictive accuracy over time and surfacing relationships that were not visible in the historical data that engineers used to design the process.

The Scaling Challenge

The manufacturers that have deployed live digital twins in one facility or one production line are now confronting the challenge of scaling across their operations. The technical challenge is manageable — the architecture that works for one facility can generally be replicated. The organisational challenge is harder: the change management required to integrate digital twin insights into operational decision-making processes, and the capability development required to ensure that operators and managers can use the tools effectively.

The manufacturers that are scaling most successfully are those that have invested as much in the human side of the deployment as in the technical side — training operators to interpret and act on digital twin insights, redesigning decision-making processes to incorporate those insights, and building the feedback loops that allow the digital twin to improve based on the decisions that operators make. Technology without adoption is not a capability — it is an asset that is not being used.

The Urban Energy Transformation

Cities are the primary arena of the energy transition. They account for the majority of global energy consumption, they are where the electrification of transport and heating is happening fastest, and they are where the tension between energy affordability, reliability, and sustainability is most acute. The energy systems that power cities are being redesigned — not incrementally, but fundamentally — and the decisions being made now will shape urban energy costs and reliability for decades.

The cities that are managing this transition most effectively are those that have developed integrated energy strategies — strategies that address generation, distribution, storage, and demand management as a system rather than as separate infrastructure decisions. The cities that are struggling are those that are making individual infrastructure decisions without a coherent view of how they fit together.

Distributed Generation as the New Baseline

The economics of distributed generation — rooftop solar, community solar, and small-scale combined heat and power — have improved to the point where they are competitive with centralised generation for many urban applications. The cities that have recognised this are building distributed generation into their energy planning rather than treating it as a supplement to centralised supply.

The strategic implication is significant. A city that has built a substantial distributed generation base has a more resilient energy system — one that is less vulnerable to transmission failures and centralised generation outages — and a lower long-run cost of energy, because distributed generation avoids the transmission and distribution costs that make up a substantial portion of the delivered cost of centralised power.

Smart Grid Infrastructure as the Enabler

Distributed generation creates complexity that centralised generation does not. Managing a grid with thousands of small generation sources, variable output, and two-way power flows requires a level of real-time monitoring and control that traditional grid infrastructure cannot provide. Smart grid technology — advanced metering, distribution automation, and grid management software — is the enabler that makes distributed generation manageable at scale.

The cities that have invested in smart grid infrastructure are finding that it delivers value beyond enabling distributed generation. It improves the reliability of the distribution system by enabling faster fault detection and isolation. It reduces the cost of peak demand management by enabling demand response programmes. And it provides the data infrastructure that allows utilities and city planners to make better decisions about where to invest in grid capacity.

The First-Mover Advantage

The cities that are moving earliest on distributed generation and smart grid infrastructure are building a cost advantage that will compound over time. The infrastructure investments they are making now will be amortised over decades, and the operational efficiencies they are building will persist. The cities that delay these investments will face higher costs when they eventually make them — both because the scale of the transition will be larger and because the window for the most favourable financing conditions may have closed.

The first-mover advantage in urban energy is not just about cost. Cities that have built modern, flexible energy infrastructure are better positioned to attract the businesses and residents that are making location decisions on the basis of energy reliability and sustainability. In a world where energy is an increasingly important input to economic competitiveness, the cities that have invested in their energy infrastructure will have a structural advantage over those that have not.

Where AI Is Delivering in Healthcare

The clinical applications of AI in healthcare are moving from research to practice faster than most health system leaders anticipated. Diagnostic imaging AI that can identify abnormalities with accuracy comparable to specialist radiologists is being deployed in radiology departments. Predictive models that identify patients at risk of deterioration before clinical signs are apparent are being used in intensive care units. Natural language processing tools that extract structured information from clinical notes are reducing the documentation burden on clinicians.

These are not marginal improvements. In each case, the AI application is addressing a genuine clinical or operational problem — diagnostic accuracy, early intervention, administrative burden — and delivering measurable improvement. The health systems that are generating these results have invested not just in the AI technology but in the data infrastructure, clinical workflow integration, and change management that make the technology effective.

The Data Infrastructure Prerequisite

The health systems that are generating the best results from clinical AI have one thing in common: they have invested in data infrastructure before deploying AI. This means electronic health records that are structured and coded consistently, imaging archives that are accessible and well-organised, and data governance frameworks that allow clinical data to be used for AI development while protecting patient privacy.

This investment is not glamorous and it is not fast. Building the data infrastructure that clinical AI requires is a multi-year programme that competes for resources with more visible clinical investments. But the health systems that have made this investment are finding that it pays dividends beyond AI — better data infrastructure improves clinical decision support, population health management, and operational analytics as well as AI model performance.

Clinical Workflow Integration

The most common failure mode in clinical AI deployment is building a model that performs well in validation but is not used in practice. This happens when the AI output is not integrated into the clinical workflow — when the clinician has to leave their normal working environment to access the AI recommendation, or when the recommendation is presented in a format that does not fit the clinical decision-making process.

Effective clinical workflow integration requires deep engagement with the clinicians who will use the AI tool — understanding how they make decisions, what information they need and when, and how the AI output can be presented in a way that enhances rather than disrupts their workflow. This is a design challenge as much as a technical one, and the health systems that have invested in clinical co-design of AI tools are seeing much higher adoption rates than those that have deployed tools designed primarily by technologists.

The Change Management Imperative

Clinical AI deployment is a change management challenge as much as a technology challenge. Clinicians who are asked to change their practice based on AI recommendations need to understand how the AI works, what its limitations are, and how to interpret its outputs. They need to trust the technology enough to act on its recommendations, but not so much that they override their own clinical judgment when the AI is wrong.

Building this calibrated trust requires investment in clinical education and engagement that most health systems have underestimated. The health systems that are managing clinical AI adoption most effectively are those that have treated clinician engagement as a programme in its own right — with dedicated resources, structured communication, and feedback mechanisms that allow clinicians to report concerns and see that those concerns are being addressed.

Aviation as an AI Laboratory

International aviation is one of the most operationally complex industries in the world. The coordination required to move aircraft, crews, passengers, and cargo across a global network — subject to weather, regulatory constraints, mechanical issues, and demand variability — is a problem of extraordinary complexity. It is also, for that reason, one of the most fertile environments for AI application. The operational gains available from better decision-making in aviation are large, and the industry has been investing in AI tools to capture them.

Agentic AI — AI systems that can take sequences of actions autonomously to achieve a defined goal — is beginning to move from research to operational deployment in aviation. The applications that are furthest along are those where the decision space is well-defined and the consequences of errors are recoverable: schedule optimisation, crew pairing, and ground operations coordination. The applications that are more nascent are those where the decision space is less defined and the consequences of errors are more severe: real-time disruption management and safety-critical operations.

Dynamic Scheduling and Network Optimisation

Airline scheduling is a classic combinatorial optimisation problem — finding the assignment of aircraft and crews to routes that maximises revenue and minimises cost subject to a large number of constraints. Traditional optimisation tools have been used for this purpose for decades, but they are typically run periodically and produce static schedules that are then managed by human dispatchers as conditions change.

Agentic AI systems are beginning to change this model. Rather than producing a static schedule that humans then manage, these systems continuously monitor the state of the network and propose adjustments in real time — rerouting aircraft to avoid weather, reassigning crews to cover disruptions, and adjusting departure times to improve connection performance. The airlines that have deployed these systems are finding that they can recover from disruptions faster and with less passenger impact than those relying on human dispatchers working with traditional tools.

Disruption Management at Scale

Disruption management is where the value of agentic AI in aviation is most visible. When a major disruption occurs — a hub closure, a severe weather event, a fleet grounding — the number of decisions that need to be made simultaneously exceeds the capacity of human operations teams. Aircraft need to be repositioned, crews need to be reassigned, passengers need to be rebooked, and all of these decisions are interdependent.

Agentic AI systems can manage this complexity in a way that human teams cannot. They can evaluate thousands of possible recovery scenarios simultaneously, identify the solution that minimises total disruption cost, and begin executing it before human operators have finished assessing the situation. The airlines that have deployed these systems are not just recovering from disruptions faster — they are recovering in ways that are more equitable to passengers and more efficient for the airline.

Implications Beyond Aviation

The lessons from agentic AI deployment in aviation are relevant to any industry that involves complex, real-time coordination across a large network of interdependent assets and decisions. Logistics, energy grid management, financial market operations, and healthcare scheduling all share the fundamental characteristics that make aviation a productive environment for agentic AI: high complexity, time pressure, and large gains available from better coordination.

The governance lessons are equally transferable. Aviation has developed a sophisticated framework for managing the introduction of automation into safety-critical operations — one that emphasises gradual deployment, extensive testing, clear human override protocols, and continuous monitoring of system performance. As agentic AI moves into other industries, the governance frameworks that aviation has developed will be a valuable reference point for organisations trying to deploy these systems responsibly.