Square Enix Veterans Say AI Cannot Replace the Human Leap From Idea to Game

Artificial intelligence is rapidly becoming one of game development's most divisive and consequential technologies. Studios are considering it for everything from early concept work and dialogue assistance to programming support, testing, localization, and asset generation. Yet for veteran developers behind some of Square's most formative role-playing games, the conversation is not simply about whether AI can make production faster. It is about what happens to creativity when tools become more capable--and what human developers must still bring to the process.

Hiromichi Tanaka, Kouichi Ishii, and Yoshinori Kitase recently looked back on their work during Square's Super Nintendo period in a discussion held for the 35th anniversary of the Mana series. Their memories covered the technological constraints, shifting plans, and smaller teams that shaped games such as Secret of Mana and the early Final Fantasy titles. But their reflections also turned toward modern production, where established engines and AI-assisted tools could give far smaller teams capabilities that once belonged only to large studios.

The trio's overall position was cautiously optimistic. AI may prove valuable in development, especially as a way to speed up prototyping and help creators test ideas. However, they believe the decisive act of bringing a new idea into existence--the journey from "0 to 1"--cannot be handed over to a machine. Tools can support a vision, but people must originate it, shape it, and ultimately decide why it matters.

A Mana Anniversary Look Back at an Era of Change

The developers' discussion naturally returned to the unusual development history surrounding Secret of Mana. The game was initially intended for Nintendo's ultimately canceled CD add-on for the Super Nintendo. That plan would have allowed for a different scope and direction than the cartridge-based release players eventually received.

For useful background on this topic, read PS5 hypervisor exploit Bounty Sparks Homebrew Clash Over AI, Linux, and GTA 6.

Akira Toriyama, the celebrated creator of Dragon Ball, had been expected to contribute character designs to that version of the project. That collaboration did not go ahead after Secret of Mana was reworked as a standard Super Nintendo game. Toriyama did, however, go on to contribute to Chrono Trigger, a project that emerged after the change in direction for Secret of Mana.

The story illustrates how volatile development could be even during an era often remembered for comparatively straightforward production pipelines. Hardware plans could vanish, projects could be rebuilt around new constraints, and teams needed to adjust quickly. Tanaka, Ishii, and Kitase noted that games of the era were generally made by smaller groups over shorter development cycles than many current blockbuster productions. That gave creators room to change course more readily.

Those changes had consequences far beyond a single title. The uncertainty surrounding the SNES CD project, together with Nintendo's decision not to move to disc-based media with the Nintendo 64, contributed to the path that led Final Fantasy VII to become a PlayStation exclusive. The result was one of the most consequential platform shifts in console gaming history, helping establish Sony's original PlayStation as a major destination for role-playing games.

Kitase was central to several of the games that defined that transition. He directed Chrono Trigger and later directed major Final Fantasy entries including Final Fantasy VII, Final Fantasy VIII, and Final Fantasy X. He was also one of the producers of the original Kingdom Hearts. His perspective on modern tools comes from someone who has watched game development move from highly specialized, tightly constrained hardware work to the expansive productions of today.

AI as a Tool, Not a Substitute for Creation

Kitase sees AI as part of a longer history of technology changing who can make games and how they make them. Modern engines such as Unreal Engine and Unity already lowered a major barrier by supplying a broad technical foundation. Developers no longer necessarily need to build every engine-level system themselves before they can begin designing a game.

That accessibility has created more opportunities for inexperienced creators to complete projects and gain practical development experience. AI could extend that trend further, potentially allowing smaller teams to attempt work that would once have been out of reach. In that sense, the technology may be particularly useful not because it eliminates creators, but because it gives creators more chances to experiment.

Kitase nevertheless stressed that there is a fundamental part of development no pre-built engine or generative system can meaningfully replace. A finished game begins with a decision: an original concept, a direction, a feeling worth pursuing, or a problem the team wants to solve. Someone has to make that first leap, then recognize which possibilities are worth turning into a complete experience.

"When it comes to creating something and taking it from 0 to 1, you absolutely need people," Kitase said. "That will never change, so the next generation of developers must have the ability to make something from 0."

That distinction is important in the current AI debate. A system may be able to quickly generate variations based on a prompt, summarize existing material, or produce a rough prototype from specified inputs. But deciding what the game should be, what it should communicate, and what compromises should or should not be accepted remains creative leadership. The veterans' comments suggest that the more convenient production becomes, the more valuable that judgment may become.

Faster Prototypes Could Mean Better Iteration

Ishii, who is strongly associated with the creation of the Mana series, also sees reasons to be excited by AI's potential. His focus is on prototyping: if ideas can be assembled and tested more quickly, teams may have extra time to revise them. Rather than accepting the first workable version of a mechanic or environment, developers could evaluate alternatives and ask whether a completely different approach would better serve the game.

For a medium built around iteration, that possibility is significant. A good game often emerges not from a single first draft but from repeated cycles of testing, criticism, adjustment, and occasional reinvention. Faster early work could leave room for more of that process, especially for teams operating under limited time and resources.

Still, Ishii warned that a polished result is not necessarily a meaningful one. If creators rely on AI alone, he argued, the outcome could look complete while feeling emotionally distant. A game's human touch is not merely an aesthetic flourish added at the end. It can be present in the unusual choices, personal priorities, imperfections, and moments of restraint that make a work feel like it was made by people with a point of view.

"Humans will need to still define the beginning and end of the working process," Ishii said. "If we rely solely on AI, the result may appear polished, but it may feel cold and lack the needed human touch."

This concern reaches beyond game development. Generative tools can make it easier to produce a large quantity of material, but volume and finish are not the same as authorship. In a game, that difference can appear in level pacing, narrative rhythm, visual motifs, player feedback, or the way a small mechanic supports a larger emotional theme. These are not necessarily tasks that can be solved simply by generating more options.

The Personal Test of a Human Touch

Kitase offered a personal example that captures why the issue is not purely technical. About two years ago, he considered using AI assistance to draft a speech for his eldest son's wedding. He ultimately decided against it because it did not feel appropriate and wrote the speech himself.

The anecdote is simple, but it frames the developers' argument clearly. A wedding speech can follow familiar structures, and a tool might produce a grammatically effective draft. Yet the value of the speech comes from the speaker's relationship, memories, and responsibility to express something personal. For Kitase, that made outsourcing the words feel wrong.

Games are obviously collaborative works rather than individual speeches, but the underlying principle still applies. Developers use tools at every stage of production, from engines and art software to audio middleware and version-control systems. The question is not whether tools belong in creative work. They always have. The question is where creators retain ownership of the intent behind the work, and whether the final game reflects genuine decisions rather than an automated approximation of them.

A Broader Shift in Japan's Development Conversation

The comments arrive as the game industry in Japan appears to be moving from an early argument over whether AI should be used at all toward a more practical debate about responsible implementation. Recent surveys have indicated that more than 85 percent of developers in Japan are already using AI technology or considering its use.

That does not mean there is agreement about every application. AI use raises questions about labor, consent, training data, ownership, quality control, disclosure, and player trust. Even developers interested in the efficiency benefits will need to determine which tasks are appropriate for automation and which require direct human authorship.

The views shared by Tanaka, Ishii, and Kitase align with recent thoughts from Nintendo veteran Shigeru Miyamoto, who has also considered how new technology could alter the role of creators. The common thread is not an outright rejection of AI. It is a belief that technological progress should expand the ability to make meaningful work rather than reduce creation to the output of a tool.

For emerging developers, that may be the most useful lesson from veterans of the Super Nintendo and PlayStation eras. Learning modern engines, production software, and AI-assisted workflows can make a creator more capable. But technical access does not automatically produce a game with an identity. The person or team making it still needs to imagine something worth building, make choices that no tool can make on their behalf, and carry that idea from its first spark to a finished form.

AI may help make the path shorter. It cannot determine where the path should lead.