By Berend Booms, Associate Editor | Future of Assets

One of the things I have always found interesting about technology is that some of its most important developments surface in places where you would least expect them. Over the past year, much of the conversation around artificial intelligence has focused on emerging agentic capabilities, embedded enterprise software, productivity tools, and increasingly sophisticated language models. Most discussions have centered on how AI will displace or augment decision-making, what this means for knowledge work, and how we can apply automation sensibly and responsibly. Considerably less attention has been given to the role computer games may play in shaping how people learn to interact with these new technological advancements. That is why Krafton’s recent collaboration with NVIDIA in order to add an AI teammate to their battle royal classic ‘PlayerUnknown’s Battlegrounds (PUBG)’ caught my attention. 

I have played PUBG on-and-off for close to 10 years. It has helped define the genre of battle royale shooters, where up to one hundred players parachute onto an island. There, they scavenge for weapons and supplies while attempting to survive longer than everyone else. The game's ‘safe zone’ decreases in size over time, which forces surviving players into more frequent and more intense encounters.  

At first glance, the addition of an AI teammate to a survival shooter feels like little more than a novelty. Players enter a match accompanied by an AI companion named Ella, who can communicate through voice, discuss tactics, identify threats, and participate in the game alongside them. She helps scavenge for weapons, equipment, and supplies, provides covering fire when required, informs the player of the direction the safe zone is shifting, and is even able to revive you, should an opposing enemy get the better of you in a gunfight.  

The reaction of the community on this new AI-powered feature is really interesting. Most players show an initial amazement at what the AI teammate is able to do. Ella can discuss where to land, call out enemies, warn players about potential dangers in the vicinity, respond to questions, and contribute to a shared objective and gameplan. She retains information from previous engagements such as the player’s name, adapts to parts of the conversation, and attempts to participate as a member of the squad. At the same time, she becomes confused, misjudges situations, provides contradictory information, and occasionally displays a level of confidence that exceeds her actual understanding of the game. 

Does AI Meet the Human Standard? 

What struck me most was how quickly players stopped evaluating Ella as a piece of software and started evaluating her according to human standards. After the initial novelty of having Ella call you by whatever name you feed her wore off, they challenged her recommendations, questioned her judgment, tested her knowledge, and occasionally became frustrated when she failed to perform as expected. Some appeared genuinely entertained by the experience, while others seemed visibly uncomfortable. Several oscillated between treating her as software (strictly non-human) and treating her as a teammate (distinctly human). 

I think this reaction is more significant than it first appears: people rarely build relationships with software, dashboards, databases, or search engines. We will rely on them heavily, but we generally appreciate and approach them as tools. Ella occupies a more ambiguous space. She communicates, participates, responds, and occasionally surprises. Whether she does these things particularly well is almost secondary to the fact that the interaction begins to resemble collaboration. 

Many of the concerns being raised about AI teammates in gaming mirror broader conversations already taking place elsewhere. Some players welcome the idea of an AI companion because it fills a gap, and allows more players to enjoy the experience who would otherwise not be able to play: not everyone has friends available to play at the same time, or for the same duration. Importantly, not everyone is comfortable interacting with other human players due to the high levels of unpredictability or toxicity that can sometimes accompany online play. In this regard, an AI teammate offers something human teammates often cannot: constant availability, predictable behavior, and seemingly endless patience – the trade-off being that the level of play these AI teammates offer in no way rivals that of skilled human gamers. 

When Software Starts to Feel Human 

While my days of spending entire weekends playing PUBG are largely behind me, I do still enjoy keeping up with developments in the game, and sometimes tune in to other players streaming their gameplay. After watching a number of players interact with Ella, I started to think about what the implications would be of having an AI teammate join you in asset management. 

In many ways, PUBG’s AI teammate is a great example of a broader shift that is beginning to emerge across many industries. For decades, technology has largely existed as something people interacted with. Increasingly, it is beginning to behave like something people collaborate with. The discussion occasionally overlaps with what has become known as the "dead internet theory," the idea that online spaces are increasingly populated by algorithms, bots, synthetic content, and automated interactions. While many of the more extreme interpretations of that theory are difficult to take seriously, I do think it captures a broader discomfort that many people feel. As artificial intelligence becomes more capable of participating in activities that were once exclusively human, the line between interaction and simulation becomes increasingly difficult to define. 

If we bring the discussion back to maintenance and asset management, this idea of ‘humanness’ becomes particularly interesting. For several years now, workforce continuity has been one of the recurring themes in conversations across asset-intensive industries. The most experienced technicians are retiring, and skilled labor remains difficult to attract and retain. Organizations are increasingly aware that much of their operational knowledge resides in people rather than systems. When those people leave, a considerable amount of experience often leaves with them. 

Many of the challenges associated with this transition have less to do with labor capacity than with knowledge capacity. Organizations can document procedures. They can capture work instructions. They can build knowledge bases and digital repositories. What remains much more difficult is preserving the context that sits around that information. Experienced technicians often recognize subtle patterns that never appear in a manual. Reliability engineers develop intuition through years of observing how assets behave under changing operating conditions. Planners learn which risks deserve attention because they have lived through the consequences of getting those decisions wrong. The challenge extends beyond preserving information: what organizations are ultimately trying to preserve is judgment. 

The Challenge of Preserving Judgment 

Viewed through that lens, the idea of an AI teammate becomes much more interesting. Imagine a technician carrying out an inspection route. An AI system listens through a headset, accesses asset history, compares current conditions against historical performance, identifies relevant failure patterns, and surfaces information that would otherwise require navigating multiple systems. As work progresses, observations are documented automatically, relevant records are updated, and potential follow-up actions are identified. 

Many of these capabilities already exist in some form today. What feels different is the degree to which these capabilities are beginning to converge. Historically, maintenance technology has followed a relatively consistent trajectory. Paper systems gave way to CMMS platforms. CMMS evolved into broader Enterprise Asset Management systems. Mobile applications brought information closer to the point of execution. Predictive maintenance introduced new forms of visibility into asset health and failure risk. More recently, generative AI has made it possible to interact with information through natural language. Each stage moved technology closer to the work itself. 

The emergence of AI teammates feels like a continuation of that progression. Information becomes easier to access, context becomes easier to retrieve, and systems become increasingly capable of participating in work rather than simply storing data about the work afterwards. 

Capability, Accountability, and Responsibility 

At the same time, I think there is a risk in becoming overly focused on capability while paying insufficient attention to judgment. Now that this functionality is live for many players, they quickly learn that confidence and correctness are not the same thing. Ella often sounds convincing, but she often makes mistakes. In a gaming environment, those mistakes are largely inconsequential – at worst, you lose the match and you queue for the next. In industrial environments, the consequences are much more severe. 

This is something I have found myself returning to repeatedly in conversations around artificial intelligence. As systems become more capable, the discussion gradually shifts away from technical feasibility toward accountability. Questions about whether AI can perform a task eventually give way to questions about when it should perform that task, how much authority it should possess, and where responsibility ultimately resides. 

Those questions feel particularly relevant in maintenance and asset management, because these disciplines are fundamentally concerned with stewardship. Decisions made in these environments affect reliability, safety, performance, and continuity. They carry consequences that extend beyond a single asset or a single site. Data, analytics, and artificial intelligence can strengthen decision-making considerably, but responsibility remains attached to the people making those decisions. 

This is also where I think the comparison with gaming is most accurate. PUBG's AI teammate offers a highly visible demonstration of a future interaction model that is already beginning to emerge across other industries. Instead of opening software, people converse with it. Instead of searching for information, people receive contextual guidance. Instead of interacting with a static system, people engage with something that appears capable of participating in the activity itself. 

The Future of Gaming – The Future of Asset Management? 

Whether these systems ultimately become trusted colleagues, highly capable assistants, or something in between remains difficult to predict. What feels easier to observe is the direction of travel. Many of the underlying capabilities are already emerging around us. Memory, contextual reasoning, natural conversation, planning, and task execution are increasingly appearing within the same systems. The question facing asset-intensive organizations may therefore have less to do with whether these technologies arrive and more to do with how they are integrated into environments where judgment, accountability, and human expertise continue to matter enormously. 

The more I think about PUBG's AI teammate, the more it feels like an early indication of where our relationship with technology may be heading. Gaming has often provided a preview of broader technological shifts. Online communities emerged there before they became commonplace elsewhere. Virtual economies appeared long before organizations began discussing cryptocurrencies. AI teammates may prove to be another example. 

If that turns out to be the case, the most important conversations are unlikely to revolve around the technology itself. They will revolve around how people work alongside it, how expertise is preserved, how trust is earned, and how human judgment continues to shape decisions in increasingly intelligent environments. Those feel like questions worth paying attention to because they extend far beyond gaming. The technologies themselves will continue to evolve. How organizations choose to integrate them into environments that depend on expertise, accountability, and human judgment may ultimately prove far more important.