Supercent retools publishing with 200+ AI agents and PSL
Key takeaways
- Supercent runs 200+ AI agents across publishing stages
- AI raised monthly games reaching profitability by tens of times within one year
- Company reports 1.8B cumulative downloads and 100M MAU
- 27 titles exceed 10M downloads; Sensor Tower ranked #7 in 2025
Supercent told Playtime Seoul that it has rebuilt its publishing operations around AI, running more than 200 AI agents across development, art, marketing and data analysis. Director Kwon spoke onstage at the Google Play event in Coex, Seoul, in a session titled “AI-changing mobile app/game ecosystems,” joined by ScatterLab CEO Kim Jong-yun and hosted by science communicator Gwido. Kwon said the biggest impact of AI is not that one person can work faster, but that the total number of business opportunities a company can try has changed completely.
Kwon described how Supercent shifted from chasing a single breakout hit to a fast-iteration model: the team ships many small games, measures real market response, and folds those results into the next project. The company credits AI with a dramatic change in throughput — after adopting AI, the number of games reaching monthly profitability-stage rose by “tens of times” within a year. That growth, the company said, comes from using AI to increase the number of experiments rather than to cut headcount.
The operational backbone for those experiments is Supercent’s in-house publishing system, Publishing System Loop (PSL). PSL defines goals and tasks for each stage of publishing so Supercent could slot AI agents into roles that had previously been human jobs. As the company sped up individual tasks, new bottlenecks appeared where human teams had to pick up outputs, so Supercent reworked workflows around AI: now non-developers also build the internal tools they need, and job boundaries have been flattened to reduce handoff friction.
AI also runs parts of live operations. Supercent’s games have an AI process that checks performance metrics every three days, makes balance adjustments within preset ranges, and validates those changes; if metrics improve, the AI’s tweaks are applied without users needing to wait for a full client update. Kwon emphasized the user-facing criterion: the count of technologies inside a game matters less than whether players experience smoother, more fun gameplay. He additionally warned that as AI lowers content production costs, capturing user attention becomes harder, and stressed the importance of making created content chosen and grown rather than just produced.
Beyond games, Supercent has been researching China’s AI-driven short-drama market since 2025 and will soon unveil its first short-drama business; the plan includes launching both a studio and a platform to extend its publishing play into video. Kwon said that pre-AI success and failure data remained an asset the company could feed into AI models so those models could operate immediately inside PSL. Supercent also reported scale milestones: five years after founding it has reached 1.8 billion cumulative downloads, 100 million monthly active users, and 27 titles with over 10 million downloads; Sensor Tower ranked Supercent seventh among global mobile game publishers by downloads in 2025, the company noted as the first Korean firm to make the top 10.
Why it matters
Takeaway: Supercent has wired AI into both pipeline and ops — PSL plus 200+ agents has multiplied how many projects reach profitability each month, and the company plans to publish AI-made games and short dramas on that same system so generated content can be operated and monetized at scale.
Original source
Google News: Inven Korea