Wargaming and Conversational AI
Following on from my posts on Synthetic Media Wraps for Manual Wargames and Matrix Game Simulations, this is a slightly edited version of Annex A of my Military Applications of Conversational AI white paper where I provide a condensed but broader overview of the potential applications of conversational AI to wargaming.
There have been a number of studies recently which have looked at the potential role for AI within a wargaming context, including those by RAND (Geist et al., 2024) and the Centre for Emerging Technology and Security (CeTAS) at the Turing Institute (Knack & Powell, 2023). The diagram above (Burden, 2025) is typical, showing the different components of a professional wargame, and where AI may have a role.
AI has a potential role to play in all of the elements of a wargame. The only reason why “human player” is shown as the only human-only role is as without a human player the event becomes a simulation not a wargame. This is not to say that the wargame can’t include AI participants and players, just that to be considered a wargame by most definitions it must include at least one human play.
The following provides a brief overview of the potential (and actual) roles that a conversational AI can play in a wargame. I am actively exploring some of these, and keenly interested in them all.
Wargame Design
Probably the hardest area for AI to contribute to, but a conversational AI as a research and design assistant, and to help generate textual and media content, and even draft rules, is already viable. The AI could also help in extracting wargame requirements from the sponsor, and compare them to its knowledge of existing wargames and wargaming techniques.
Scenario Design
Given a set of rules and a scenario template, a conversational AI can readily generate alternative scenarios, and potentially the media assets required to go with them. I used AI to generate a consistent set of images to illustrate the capability cards in The Urban Calculus: Evolved, and to generate the city briefings in some City & CEMA scenarios.
Scenario Setup
Setting up a game can often take time, and it’s easy to miss things. A chatbot can help guide the players through this, and answer any questions on setting up the game.
Player
Conversational AIs are increasingly able to digest a set of rules and to become players in a game. For a manual game they can use a human as their “agent” to describe the game state and options available, and to take the physical actions to make a move. Of course, a wargame devoid of any human players is a simulation not a wargame, but an AI player can provide an alternative to solo play, or fill in if players are missing, or specialize in playing the adversary and reflect the adversaries tactics and doctrine. My Matrix Game Simulation has all the players as an AI (and so is a simulation not a wargame!). The role of AI as the opposing player is of course of keen interest, and I’m slowly working to develop one for some of my hobby and PhD games.
NPCs
Chatbots are more than capable of taking on the role of non-player characters (NPCs)– such as allies or intelligence sources, and a lot of earlier conversational AI work was driven by creating more believable NPCs in sandbox computer games.
Player Support
A chatbot can help a player understand the rules, answer questions on them, and advise on game strategy and tactics. My Virtual Urban Wargamer chatbot has all of my wargame rules and so can perform this role for each of my games, and I’m currently looking at how to optimise bots for this role.
Umpire/ Adjudicator/ Facilitator
Another major challenge area for chatbots, but LLMs are already being used to umpire conversational based wargames such as matrix games, and could increasingly be used to help facilitate or adjudicate where an umpire familiar with a specific game is unavailable. Again my Matrix Game Simulation has the bot perform this role, as well as that of the player, and I’m also using ChatGPT to act as games-master in Role-Playing Games.
Umpire Assistant
Umpires have a lot to contend with in a game, particularly if covering umpire, facilitation and adjudication functions. Chatbots could support an umpire across all of these, helping with rules questions and executions, and potentially being able to take a step back and help with teaming and focus issues.
Data Collector
Professional wargames typically have a well-developed data collection plan, which may include scribes/observers and even question probes. Chatbots can readily augment or provide much of this support, enabling more players to be interviewed in depth about their experiences with the game. Duchess is a commercial example of a bot developed for operational review collection but which could just as easily be applied to wargames. A past project I did for MOD used De Bono’s Six Thinking Hats as a basis for an interview bot, and we also had a bootstrap bot in the Virtual Humans project, both of which I’m porting to current LLMs.
Analyst
Whilst non-conversational AI probably takes the lead in analysing the wargame output, much of the output is on the qualitative side which will require natural language understanding, and conversational AI can act as an analysts assistant in managing the workflow and helping to link findings to the literature. As a test I recently had an LLM analyse the audio transcripts of AARs after one of my wargames, and the output was very close to what I produced manually.
This is just a very quick survey of potential and actual roles, and as AI capabilities continue to develop I am sure that we will see more and more use of AI to support wargames, and to run simulations. One key element for me is how we democratise this approach, so just as almost anyone can design a wargame almost anyone can also make use of AI tools to support it, and they don’t become the sole preserve of “big tech” and the large defence primes.
References
Burden, D. (2025). Wargaming and the Just War. In M. Power & M. Savin-Baden (Eds.), Just War Theory and Artificial Intelligence. Taylor & Francis. https://www.taylorfrancis.com/books/edit/10.1201/9781003440253/war-theory-artificial-intelligence-maria-power-maggi-savin-baden
Geist, E., Frank, A. B., & Menthe, L. (2024). Understanding the Limits of Artificial Intelligence for Warfighters: Volume 4, Wargames. RAND Corporation. https://www.rand.org/pubs/research_reports/RRA1722-4.html
Knack, A., & Powell, R. (2023). Artificial Intelligence in Wargaming. CETaS, Alan Turing Institute. https://cetas.turing.ac.uk/publications/artificial-intelligence-wargaming
You can download and read my full white paper on the Military Applications of Conversational AI.
