Kewpie has announced that it conducted a verification project to reduce the development workload of factory simulation models by utilizing the autonomous AI agent "Devin." Supported by DeNA AI Link, the company verified the efficiency of model development within Siemens' simulation software, "Plant Simulation."
Kewpie reduces factory simulation model development workload by 78.8% using AI agent Devin
This article is a translation. Read the Japanese original
Constructing simulation models requires not only the operation of specialized software but also advanced knowledge of the domain-specific language "SimTalk," which has made individual dependency a challenge. In this initiative, the company used the cloud version of Devin for knowledge sharing and infrastructure building, while utilizing "Devin CLI" for execution and modification in local environments, establishing a system where the AI autonomously executes and verifies code.
As a result of the verification, the development workload per 1 model—which was previously estimated to require 8 person-days of manual work—was reduced to 1 person-days, confirming a 78.8% reduction in workload. Moving forward, Kewpie aims to expand the scope of supported simulations and realize a system where on-site personnel can construct models using natural language.
Sources
- キユーピー、AIで工場シミュレーションモデルの開発工数を8割削減 3つの壁をどう乗り越えた？ （ITmedia AI+、2026-09-22）
- DeNA