In applying AI to military decision-making, the IP4 should learn from NATO

NATO’s Indo-Pacific Four (IP4) partners – Australia, Japan, South Korea and New Zealand – left the alliance’s July summit in Ankara promising more collaboration on technology development, equipment production and procurement processes. With all IP4 members interested in integrating AI into their militaries, their first order of business should be learning from NATO’s rapid adoption of AI-enabled decision-support systems (AI-DSSs).

Such systems are software that ingest, analyse, visualise and share vast and disparate data to support more effective decisions. Their benefits are being felt across the targeting cycle, including more effective target nominations against official standards, language translation and formatting of targeting packages (electronic folders detailing target information). These benefits stand to alleviate some coalition targeting challenges, such as achieving consensus on which targets to strike and determining which nations are capable of striking them. Operation Allied Force in Kosovo, for example, required the unanimous approval of 19 nations for each strike. Broadly, AI-DSSs are being used to augment human ability: users have said the technology is helping with task overload and is flagging errors with human-generated content, such as briefings, assessments, reports and data.

In March 2025, NATO’s Supreme Headquarters Allied Powers Europe (SHAPE) announced the acquisition of the best-known American AI-DSS, the Maven Smart System. Taking only six months from the requirement outline, the acquisition was the fastest in NATO’s history, which should give IP4 members cause for optimism. A little more than a year on, NATO Maven is fully authorised for operational use and security-accredited to operate on classified NATO networks. Already, Task Force Maven – the contingent within NATO SHAPE responsible for the alliance’s adoption of Maven – has reported $6 million in annual savings from the system’s use in training. Its benefits have impressed NATO members such that 10 of them have reportedly lined up to buy the technology independently.

NATO’s quick adoption of the Maven Smart System should come as little surprise when considering the strengths of AI-DSSs against the difficulties of coalition operations. Take the former coalition operations in Afghanistan as an example: even the most basic details about national caveats (the rules countries applied to how their militaries might be deployed) were previously tracked in enormous spreadsheets. This made the shared information difficult to manage and vulnerable to errors that, at worst, resulted in near-fatal accidents in the field. Humanitarian operations, such as the response to the 2004 Indian Ocean Tsunami, have also proven frustratingly complex for coalitions that must manage ever-changing information in multiple languages across numerous networks and classification levels.

AI-DSSs are well suited to help with these and other coalition challenges. Whether it is sufficiently declassifying intelligence reports to share with partners or managing national caveats, AI-DSSs can help navigate nearly any process that involves interpreting data, assessing a situation, making decisions and then implementing those decisions. AI-DSSs can therefore free up time for planners to consider more difficult decisions.

Given these systems’ promise, it would be a mistake to let a learning opportunity pass. IP4 coordination on this should start from the premise that the partners each approach AI from different baselines and with varying degrees of trust in the technology itself. To overcome hesitancy, the IP4 should build confidence in AI-DSSs through demonstrations, exercises and transparency. Using the IP4 as a forum to have robust discussions about AI-DSS applications and requirements could, for example, improve readiness for future military operations. Building experience through exercises, such as AUKUS Maritime Big Play, which focuses on testing and improving technology, is another way to do this. After all, leaders from Japan, South Korea and New Zealand have all expressed interest in observing or joining AUKUS Pillar Two activities.

Progressing this would not be without friction. Partners will likely face the same challenges that have plagued every interoperability and technology integration effort over many decades. These include reconciling data, security and interoperability standards, which have proven difficult for AUKUS. However, if NATO Maven has shown anything, it’s that a crisis – in this case, the war in Ukraine – has a funny way of clearing sticky political and bureaucratic hurdles.

This article is an adapted excerpt from Coalition Operations: The Untapped Role of AI in Military Decision-Making, a forthcoming CSET report by Emelia Probasco, Lauren Kahn, and Sophie Mayo.