Warhorse CEO Says AI Cannot Replace Developers, but Could Help Teams Build Bigger Games

Generative AI remains one of the most divisive subjects in game development, and Martin Klíma, CEO of Kingdom Come: Deliverance developer Warhorse Studios, has offered a clear view on where he believes the technology does and does not belong. His central point is direct: using AI as a way to replace developers is a "fundamentally flawed idea."

That does not mean Klíma sees no potential for AI-assisted tools. Rather, he distinguishes between treating AI as an automatic substitute for human creative work and using it to make an existing team more efficient. In the latter case, he believes the technology could eventually help developers make larger games without necessarily increasing a project's cost.

It is an important distinction at a time when players, artists, writers, actors, and developers are scrutinizing every announced use of generative technology. The question is no longer simply whether AI can create an image, a line of dialogue, or a piece of audio. The more consequential question is what studios expect the technology to do, who retains creative responsibility, and whether efficiency gains are being used to support workers or reduce their role.

AI Is Not a Shortcut Through Development

Klíma rejected the idea that generative AI offers a simple answer to the complexity of making games. He described AI as a tool rather than a "get out of jail free card," pushing back on a simplified assumption that a studio can replace staff with a prompt-driven system and continue producing work at the same standard.

For useful background on this topic, read Tom Rhys Harries Used Elden Ring to Cope With Clayface Screen-Test Anxiety.

That assumption overlooks how many connected disciplines contribute to a finished game. Game development is not just the production of individual assets. It involves planning, iteration, technical implementation, quality assurance, performance considerations, narrative consistency, art direction, combat balance, quest design, accessibility, localization, and the countless decisions required to make all of those pieces work together.

Even an asset that appears usable in isolation must fit a wider production pipeline. A character concept needs to match the game's visual identity. A line of dialogue needs to work with the scene, the character, voice performance, quest states, and localization. A mission needs to account for player behavior, progression systems, rewards, bugs, and pacing. Those are collaborative and iterative processes, not problems solved merely by generating more raw material.

Klíma's comments therefore focus on the gap between a broad public perception of AI and the practical realities of building games. Replacing people with generative systems may sound like a simple cost-saving equation, but it does not eliminate the need for experienced people to define goals, evaluate results, fix failures, and make final creative decisions.

Efficiency Could Still Mean More Ambitious Games

While Klíma is skeptical of replacement rhetoric, he does see a possible benefit if AI is approached as support for a human-led development process. His suggestion is not that the technology would make games cheap or effortless to create. Instead, it could enable a team to deliver more within an otherwise similar budget.

That additional scope could take several forms. Klíma pointed to possibilities such as more branching paths and more missions for players to experience. In an RPG, those kinds of additions can have a major impact on perceived depth and replayability. More choices can create different outcomes, more reasons to revisit a quest line, and a stronger sense that the world responds to player decisions.

However, branching content is also expensive because each choice can create additional work across writing, design, testing, cinematics, voice production, and bug fixing. A tool that meaningfully reduces friction in selected parts of that process could, in theory, allow a studio to devote time and resources to more content rather than simply cutting its workforce.

"Bigger games for the same cost" is therefore the key idea in Klíma's argument. It is not a promise that AI will lower development budgets, and it is not a declaration that Warhorse will use generative assets. It is a case for considering whether carefully limited, practical applications might let teams achieve more while keeping human creators at the center of the work.

The Debate Depends on How Studios Use the Tools

That distinction matters because the industry's most contentious AI examples have often involved generated assets, including imagery, text, video, or audio. Those uses can raise concerns over creative ownership, consent, training data, visual consistency, and the replacement of professional labor. They can also provoke a player response that is less about the technology itself than about whether a company appears to be treating human-made work as disposable.

Klíma's comments suggest a narrower, process-oriented approach. Instead of asking a generative system to replace artists, writers, or other specialists, a studio might investigate tools that ease repetitive administrative burdens or speed up less creatively defining tasks. The details matter greatly. An efficiency tool that helps a team organize work is fundamentally different from a decision to remove the people responsible for the game's art, writing, performance, or design.

For a studio making a complex role-playing game, this could be especially relevant. Games in the Kingdom Come: Deliverance series depend on large environments, numerous quests, systemic interactions, historical settings, and a considerable amount of authored material. Building that type of experience requires a large web of connected work, even when a team is modest in size compared with the scale of the final game.

Using software to reduce friction around parts of production could give specialists more time to focus on the decisions that players notice most. But the value of that outcome depends on a studio's priorities. Efficiency can be used to improve quality, expand scope, protect schedules, or support employees. It can also be framed primarily as a way to reduce headcount. Klíma's criticism is directed at the latter mindset.

Human Direction Remains Central

The CEO's position aligns with a broader argument heard from prominent game creators: AI may have a role as an instrument, but it should not be treated as an autonomous replacement for the people making a game. Hideo Kojima, known for Metal Gear and Death Stranding, has similarly described AI as a tool while expressing opposition to replacing workers with it.

This perspective does not settle the debate, but it emphasizes an issue that can get lost in discussions of automation. A game's identity comes from choices: what it is trying to say, how it feels to play, which ideas are prioritized, what gets cut, and where its creators spend their limited time. Tools can assist with work, but they do not remove the need for judgment, taste, accountability, and collaboration.

It also recognizes that development efficiency is not inherently harmful. Studios have always adopted new engines, middleware, pipelines, automation, and production tools to make ambitious projects possible. The concern is whether generative AI is being introduced as another aid for creators or as a justification to eliminate them.

Warhorse's Position Comes Amid Industry Anxiety

There is no single industry-wide approach to generative AI. Different companies are experimenting in different areas, and public reactions have varied sharply. Some players are open to tools that make development smoother, particularly when human authorship remains clear. Others are wary of any use of generative systems because of unresolved questions surrounding labor, transparency, and the material used to train AI models.

Those concerns are unlikely to disappear soon. As tools improve, studios will continue to face pressure to explain how they are being used and why. A vague claim that AI will make development more efficient will not answer questions about whether jobs are protected, whether generated material appears in a shipped game, or whether the technology actually improves the player experience.

Klíma's remarks do not present AI as a miracle solution. Instead, they set a boundary: replacing developers is not a workable understanding of how games are made. At the same time, he leaves open the possibility that tools which genuinely assist teams could help developers deliver more ambitious games without requiring proportionally larger budgets.

For players, the eventual test will be visible in the games themselves and in the practices behind them. If new technology gives developers more room to create richer quests, more meaningful choices, and more polished worlds while preserving human craft, the conversation may look very different from one focused on replacing the people who make games possible.