TECH2030
Insights
Case Study · 4 min readAI-assisted

AI Production Planning APS Boosts On-Time Delivery Rate by 18.16%

The JNEWorks consortium's APS achieved a 16.13% makespan improvement and 18.16% on-time delivery boost in its 2025 evaluation.

AI Production Planning APS Boosts On-Time Delivery Rate by 18.16%

Key Takeaways

  • APS Evaluation: 16.13% Makespan Improvement
  • AI's Role: Rescheduling Tasks
  • The Key Is Connecting Equipment and Material Data
  • Start by Counting Rescheduling Frequency and Time

The autonomous production planning solution 'APS,' developed by a consortium of JNEWorks, VMS Solutions, and Changwon National University's Industry-Academic Cooperation Foundation, achieved a 16.13% improvement in makespan and an 18.16% increase in on-time delivery rate in its 2025 algorithm evaluation. In the same evaluation, Bi-LSTM-based demand forecasting accuracy reached 96.23%. The Electronic Times reported this on November 16, 2026.

Makespan refers to the total time required to complete production, and the 16.13% figure represents the improvement compared to the existing plan. APS is a solution in which AI generates a new production schedule when disruptions occur, such as urgent orders, equipment failures, or delays in material supply.

What Changes When Production Is Disrupted?

The demonstration site, Taerim Industry's Changwon plant, produces a wide variety of parts, including automotive steering components, and frequently experiences production variables such as model changes, urgent orders, and workforce fluctuations. Previously, staff had to manually revise production plans in Excel, a process that took more than half a day to adjust schedules, and it was difficult to simultaneously reflect worker schedules, equipment load, and material supply conditions.

APS combines the following three elements into a single platform.

  • Bi-LSTM-based demand forecasting AI
  • Reinforcement learning-based production rescheduling
  • 'What-if' simulation

When an urgent order comes in or a specific piece of equipment stops, the AI recalculates the existing production sequence and equipment allocation, then presents a new production plan by comparing on-time delivery rate, production time, and equipment load. If a material supply issue arises, the system can also reorder the production sequence and schedule by taking into account the tasks that require that particular material.

How Far Has the Demonstration Progressed?

In 2026, the consortium integrated these functions into an APS Software-as-a-Service (SaaS) platform that can be used across multiple processes and by multiple users. The consortium also pursued a field demonstration that connects the platform with Taerim Industry's AAS-based data lake to generate production plans based on actual production and equipment data.

The scope of application has expanded from automotive parts to shipbuilding, marine, and wind power equipment. The solution was applied at Sebo Technology in Busan, where it verified production plans responding to project-specific specification changes, delayed safety tests, and equipment failures, while connecting different manufacturing environments through standard APIs.

The consortium announced the following targets to be verified through TTA testing:

  • Production planning automation rate of 85% or higher
  • APS data integration success rate of 95% or higher
  • Makespan improvement rate of 20% or higher
  • On-time delivery improvement rate of 20% or higher
  • What-if multi-scenario evaluation time within 20 minutes

The consortium plans to commercialize the solution as a cloud service that reduces the implementation burden for individual companies, and to expand its application to small-batch, high-mix production environments and fields with frequent schedule changes, such as defense, ports, and plant engineering. This project is part of the 'Regional-Linked AI Solution Development and Demonstration Support Project (2025-2026),' led by JNEWorks as the lead company under Gyeongsangnam-do and the Gyeongnam Technopark, and is a 2026 detailed project under the 'Manufacturing AI Convergence Infrastructure Project (2024-2026)' supported by the Ministry of Science and ICT and the National IT Industry Promotion Agency (NIPA). The two-year project began in April 2025 and will complete field demonstration and performance verification in 2026.

Implications for Korean Enterprise AX

In this case, AI's role is not production itself but rescheduling—the task that used to tie up staff with Excel spreadsheets for more than half a day every time a disruption occurred. Hong Eui-seok, CEO of JNEWorks, said, 'The core of this project is reducing the burden on staff who had to redo the entire plan from scratch every time a production disruption occurred.' When evaluating production planning AI, it makes sense to look first at how often rescheduling occurs and how long each instance takes, rather than focusing solely on accuracy figures.

It's also worth noting that the Taerim Industry demonstration connected to an AAS-based data lake to build plans using actual production and equipment data. For rescheduling AI to compare alternative schedules, it needs to read worker schedules, equipment load, and material supply data from a single source. If this data exists only within a staff member's spreadsheet, connecting it becomes the first step toward adoption.

This week, try recording how many times over the past month production plans had to be rescheduled due to urgent orders, equipment failures, or material delays, and how long each instance took.

Source: JNEWorks Consortium Develops AI Autonomous Production Planning Platform...Shipbuilding and Marine Manufacturing

If you found this helpful, share it.
Related servicesAX Transformation
Enterprise AX
Want to apply this to your own operations?
Request an AX consultation