April 23, 2024

The Ministry of ChatGPT? – A quick intro to AI in Defence


AI in defence is a tricky one; in theory the real-world applications of AI will mostly be to enable humans to make better, quicker, safer decisions, and execute these more efficiently; this should lead to net better outcomes by enabling greater control over risk. However, the thought of AI in defence automatically conjures visions of killer robots, self-programmed to carry out autonomous missions, etc. etc.

What is AI?

AI is the ability to automate tasks that would typically be carried out by humans (applying our biological intelligence and intuition) – increasingly, the concept of machine learning is grouped into this, whereby the system is capable of learning from patterns and applying its learning to determine future processing and operation.

As technology evolves, so do our vulnerabilities, threats, and opportunities. AI isn’t new; algorithms have been around for years – the main difference now is the sophistication of these algos, and their ability to learn and evolve themselves without the need of a human to tweak the lines of code in the background. The recent excitement around Chat GPT combining a LLM (large language model) with natural language processing (the ability to understand “normal speak” rather than requiring us to communicate with it in keyword-jargon) has catalysed public usage of AI, and with it, the awareness of AI. Commercial vendors have started to implement LLMs into their existing propositions, offering a new axis for differentiation (through enhanced service delivery), and this trend will be seen across industries.

So, what might AI give defence?

AI gives us the ability to advance and enhance our capabilities, make better and quicker decisions, improve the efficiencies of our organisation, and remain relevant and competitive as adversaries invest in their offensive and defensive AI enabled capabilities. Some of the core principles AI can deliver are:

1.       Decision salience: Quicker, better, distributed decisions – make sure the right person has the right information, in the right location, in the right format, with sufficient time to enable an appropriate response (sometimes this may be an effect, but not necessarily)

2.       Capability advancement: Deliver enhanced capabilities, and net new capabilities across all domains. In addition to automation, and optimisation, AI can lead to step-changes in capabilities (just look within the Cyber realm to see this already at play)

3.       Efficiency: Continuous, data-led optimisation. This extends from operations through to to all levers of the organisation. Utilising personnel more efficiently, operating assets more optimally, predictive maintenance etc.

4.       Capture human value-add: Re-focus our personnel on high-value tasks by automating routine, administrative tasks. There will always be things that humans can do better – intuition, emotional intelligence, these are all things that are (yet) difficult to emulate within AI. In a future where AI becomes ubiquitous, this human value-add will become a powerful axis for military superiority. Also, manned-machine teaming should in theory deliver a greater sum effect than their individual parts

5.       Force multiplier: Deliver more from our existing capabilities, and operating models (integrating people, systems, etc.)

6.       Counter the adversary: Whatever our views on AI, adversaries are investing in it, and will continue to develop and deploy offensive and defensive measures that we need to be prepared to deal with

Lets visualise this in practice:

If we think of one potential application, the core intelligence functions or the intelligence cycle; simplistically, it comprises 5 main activities – we Direct assets to Collect intelligence, that is then Processed, Exploited, and finally, Disseminated. So, there is clear direction on what intelligence to gather, which then determines with assets are prioritised and tasked. Data is collected, processed, and value is extracted from it, that is disseminated to the appropriate function where decisions are made on what to do with this data.

The data is collected via numerous assets, and from a variety of sources, SIGINT, MASINT, GEOINT, IMINT, HUMINT, OSINT, etc. – that’s a lot of INT from a lot of sources, with varying degrees of confidence. What makes it even tougher to process and act on is the ever-changing nature of intelligence; a constantly moving picture. As the scale, and complexity of the Defence estate grows, and as multiple domains, govt. agencies, allies start generating and sharing intelligence, this moving picture becomes even more complex to manage, interpret, interrogate, and act on.

The scope for AI here is HUGE; from tasking assets (based on parameters input from humans, and the evolving intelligence picture) to optimising the deployment patterns of those assets (to ensure mission completion, coverage, through-life performance, pre-emptive maintenance, etc.) to collating and analysing the data (against the intelligence collection plan), disseminating and presenting the data to a human in the loop in a format conducive to decision making, and within certain operational contexts and thresholds, enabling additional assets (intelligence or fires) to be tasked based on this intelligence picture.

AI therefore can play into every step of the value-chain, optimising existing processing, and providing step-changes in capability, lead-times, and operational efficiency. In addition to these benefits, the humans currently focused on parts of this value-chain can potentially be tasked with performing additional higher-value tasks where biological intelligence and intuition can lead to significant comparative advantage.

Recently, Iran unleashed ~150 explosive drones, ~100 ballistic missiles, and ~30 cruise missiles on Israel in a bid to distract and overwhelm their air defences. These types attacks (similar to those deployed by Russia on Ukraine) can be thwarted by existing capabilities (as evidenced by Israel’s use of its Iron Dome, Patriot system, combat air assets, and other capabilities, alongside those of its Western and Arab allies including US, UK, and Jordan). One of the challenges with employing this type of defence is you risk using defensive measures e.g. missiles that can cost upward of $1M a pop to take down a drone worth $1,000’s (not to mention you risk running out of suitable defensive measures before the next wave of targeted missile attacks come through). In this example, AI has a lot of scope to provide greater support in the early identification and processing of hostile target information, to then support determining and tasking the appropriate defensive assets relative to the threat (taking into consideration the threat, risks, economics, and other parameters).

Where will this be applied?

The example above combines a few different threads of AI application, but is by no means exhaustive. The UK MoD’s AI playbook splits AI application into 5 areas:

1. Autonomous logistics – supporting e2e logistics and resupply

2. Exploiting operational data – enabling an automated ISR enterprise (similar to our example above)

3. Human-Machine teaming for military effect – augmenting existing force structures with trusted autonomous adjuncts

4. Machine-speed decision making – delivering quicker C2

5. Managing the defence enterprise – enabling a more efficient organisation

In the near term, we’re going to see AI solutions play out in a number of places – ISR data analysis, edge processing, strategic decision making, warfare systems, combat simulation, target recognition, threat monitoring, drone swarming, cybersecurity, transportation, etc. etc. the list goes on and on. Longer term, as the technology becomes increasingly sophisticated, who knows what might happen..

There is a lot of complexity that sits between us and a viable, operational AI enabled organisation. For an op modeler, this is a huge and exciting opportunity space.

The simplified view here doesn’t begin to touch on the risks and issues with embracing AI within this context, nor does it touch upon the governance burden this creates, or the complex ethics at play. More on this in future editions.


Anirudh is a Principal at BCE Consulting based in London UK. Anirudh’s experience spans Strategy through to Transformation, covering Corporate and Business Unit Strategy, M&A advisory, Operating Model design, Change management and complex Programme and Project management. Anirudh’s industry specialism is across the A&D supply-chain where he has worked on various Strategy and Transformation engagements for the likes of Airbus, Babcock, Leonardo and the various arms of the UK Ministry of Defence. Anirudh holds an MBA from Warwick Business School.

Global A&D

Anirudh Suneel
Principal London
Craig Belanger
Senior Partner & Co-Founder Boston
Joe Giandomenico
Principal Boston
Ben Osterholtz
Manager Boston
Robyn Pirie
Manager Boston
Mark Kipphut
Senior Advisor Dallas
  • Menlo Park, CA
  • Boston, MA 100 High St,
    Boston, MA 02110
  • Yarmouth, ME 121 Main St,
    Yarmouth, ME 04096
  • London, UK 30 Crown Place,
    London, UK

'What's next' starts here

Connect with us Explore careers