The Strategic Defence Review is explicit: technology is changing how war is fought, and the advantage will go to those who get it into the hands of their armed forces fastest. Our view goes further. AI, autonomy, software, data and digital connectivity are not simply tools for improving the defence organisation and force we have today. Used together, they can fundamentally reshape how defence senses, decides, fights, sustains itself and works with industry and allies. But that potential will be realised only if defence redesigns its operating model around the decisions and effects these technologies make possible.
We are both technologists, but we approach the challenge from complementary starting points. Ed’s perspective is grounded in defence, its missions, operating environment and the realities of delivery. Ross brings a wider technology lens, focused on how digital change is delivered and scaled, applied to the complex enterprise that puts the capability in the hands of the warfighter. We reach the same conclusion: Defence is at the beginning of a technology-led redesign of how it works and fights, not another cycle of technology adoption. The central leadership challenge is to convert that technological possibility into decision and operational advantage at greater pace than it has done to date.
The Strategic Defence Review sets the direction: a more lethal Integrated Force; a ‘NATO first’ posture; new partnerships with industry; and drones, data and digital warfare drawn from the lessons of Ukraine. It also signals a Digital Targeting Web and a force designed to evolve continuously as threats and technologies change. Those ambitions make two issues unavoidable: the speed and quality of decisions must become a defence capability in their own right, and the pace at which new technology is fielded must itself be treated as a source of operational advantage.
That is a different proposition from digitising existing processes. AI can compress planning cycles, generate and test courses of action, orchestrate sensors and effectors, and allow autonomous systems to operate at a scale and tempo that people alone cannot match. Digital twins can change how forces train and assets are sustained; software-defined capabilities can evolve between - and during - deployments; robotics can alter the balance between mass, lethality and human exposure. The opportunity is not a faster version of today’s defence. It is a different operating model and, increasingly, a different force. Decision advantage is the means by which that technological potential becomes operational effect.
For senior data and digital leaders, this shifts the agenda from programmes and platforms to three design questions:
Start with mission outcomes, not use cases. Identify the small number of decisions that govern operational tempo; readiness, targeting, threat response, sustainment, re-tasking and industrial replenishment. Then work backwards through the data, authorities, models, people and systems required to improve them. This exposes where latency really sits: often not in the algorithm, but in data ownership, assurance, hand-offs and decision rights.
AI can do more than help people interpret sensor feeds, intelligence reporting, cyber alerts and logistics signals at a scale no human team can manage alone. It can continuously fuse those signals, identify changing patterns, propose options, re-plan as conditions shift and coordinate action across domains. Combined with autonomy and software-defined systems, it can change the distribution of work between people and machines; and connect sensing, deciding and acting in ways that are impossible through traditional command and staff processes. This is not about removing human judgement. It is about placing judgement where it adds most value while machines provide speed, scale and persistence.
Behind the mission, operations and battlespace sits a huge machine whose purpose is to put capability in the hands of the warfighter at the speed of relevance. Whilst the UK delivery ecosystem it a strategic asset which has some leading capabilities and a workforce proud to serve, the collective outcome from the machine is delay, overspend and bureaucracy. The time to investment case approval is excessive and the major project report show overall programme delay has increased from 78 months 10 years ago to 300 months today. Pace is not simply a programme-management concern: in a contest where technology, tactics and threats evolve continuously. Standing still now means going backwards. A capability delivered years late does not preserve today’s position; it arrives into an environment that has already moved on.
Where we get really excited is about the opportunity to leapfrog. Defence does not have to replicate every stage of the technology journeys taken by large commercial organisations, nor replace each legacy process with a digital facsimile. It can move directly to cloud-native, software-defined and AI-enabled models; design around common data and modular interfaces; and use autonomous systems and AI agents to bypass layers of manual coordination. The constraint is less the availability of technology than the willingness to retire old assumptions, change decision rights and field usable capability before every uncertainty has been removed.
That requires a different delivery rhythm: place small increments of capability with users in weeks and months, learn through operational experimentation, and scale what works across the enterprise and force. Shared interfaces, deployable data standards and commercial routes must make successful products easier to adopt than to block. Assurance should be continuous and proportionate to risk, not a final gateway after the moment of advantage has passed. The aim is not reckless speed. It is disciplined speed, moving quickly enough to learn before an adversary does.
Trust must be engineered into that path. People need to understand the provenance of data, the limits of models and who is accountable for the outcome. Human judgement remains essential, but ‘human in the loop’ cannot become shorthand for an undefined manual checkpoint that removes the speed advantage. Defence should be explicit about where a person decides, where a person supervises and where automation is permitted within clear boundaries. Accountability and empowerment has been diluted and it's more important that this is re-established within AI based delivery models.
Warfighting readiness cannot depend on perfect connectivity, uninterrupted cloud services or a single supplier remaining available. Digital resilience must therefore be expressed in mission terms: which decisions can be delayed, which can be made locally, which require allied data, and which services must continue in a degraded mode?
This is where cyber resilience, data architecture and operational planning converge. The right question is not whether a system is secure in isolation, but whether the decision chain remains credible through attack, disconnection, data corruption or supply-chain failure. Recovery priorities should follow operational value, not application ownership.
The same transformation applies behind the frontline. AI agents could reshape headquarters and corporate functions by preparing options, coordinating workflows and maintaining a live view of readiness. Predictive systems can anticipate demand and equipment failure; digital twins can test mobilisation and sustainment choices before they are made; automation can release scarce people from routine administration. Applied across finance, HR, commercial, logistics and shared services, these technologies could also deliver the back office at materially lower cost. Simplifying demand, reducing duplication and automating high-volume work. That matters strategically: productivity gains should not disappear into general efficiency targets, but be explicitly measured and recycled into frontline capability, readiness, stocks and innovation. The prize is therefore more than efficiency. It is a smaller decision burden, a more adaptive defence enterprise and greater operational headroom, with more of the defence pound directed to fighting power.
Senior leaders should demand evidence through exercises, not assurances on slides. Test a targeting, sustainment or mobilisation decision with connectivity lost, a key dataset unavailable and a supplier compromised. The result will reveal whether resilience is real, where authority becomes ambiguous and which dependencies need redesign.
Sovereignty is too often reduced to where data sits or who owns a platform. For defence, the more useful test is freedom of action: can the organisation govern, change, recover and continue the capabilities that matter to the mission on terms it understands? Commercial lock-in, skills availability, performance trade-off and redundancy are equal considerations to ensure freedom of action.
That does not mean owning everything or retreating from commercial and allied ecosystems. NATO interoperability and access to leading technology are essential. The answer is deliberate dependency: knowing where reliance creates advantage, where it creates unacceptable constraint, and what technical, contractual and operational fallbacks are required before a crisis.
Leaders should map critical decisions to their dependencies across data, models, compute, networks, skills, intellectual property and suppliers. For each dependency, decide whether defence needs control, choice, continuity, or a combination of all three. This creates a far more practical sovereignty agenda than a blanket preference for one hosting model or national label.
The Strategic Defence Review is a mandate to change how defence fights, buys, partners and innovates and to do so at wartime pace. AI and other technologies are central to that change, not a supporting workstream. For the data and digital community, success should therefore be measured by whether technology enables a different force and a different enterprise: faster decision cycles, new combinations of crewed and autonomous systems, continuously evolving software-defined capability, more resilient sustainment, lower-cost back-office services and a workforce focused on the tasks where human judgement matters most. Speed to field, speed to learn and speed to scale must sit alongside resilience and trust as leadership measures. Crucially, the value released in the enterprise should be converted into additional frontline capability rather than treated as an abstract efficiency dividend. Platform adoption, migration milestones and model accuracy remain important, but they are intermediate measures.
Our challenge to the leaders in the room is simple: do not ask only where AI can be added to defence today. Choose one mission-critical decision, fighting function or enterprise process and ask how AI, autonomy, software and data would allow it to be designed from first principles and what could be fielded within the next 90 days. Give it a named operational owner. Measure its current latency and failure points. Redefine the roles of people and machines. Expose its data and supplier dependencies. Define how it works when degraded. Then put an initial capability in users’ hands, learn rapidly and scale it. The choice is not between moving quickly and moving safely; it is between disciplined pace and relative decline. That is how the SDR moves from technology adoption to a defence organisation and force genuinely reshaped by technology.