Modernising with Jidoka
Dr Eban Escott presented the following slides and information on Jidoka - Automation with a Human Touch at Tech In Gov. The talk focused on explaing the concept of Jidoka by introducing the concept and explainging some real work examples of it in use for Quality Control.
Jidoka, often translated as “automation with a human touch,” originates from Japan and was popularised by Toyota. While the standard term jidōka means automation (自 – self, 動 – movement, 化 – transformation), Toyota’s version replaces “movement” with “work” (働), adding the human element into the concept. First introduced through Sakichi Toyoda’s loom that automatically stopped when threads broke, Jidoka ensures machines detect abnormalities, stop processes, and allow people to focus on solving root causes. In his 2024 paper, Eban Escott extends this principle to software development, showing how continuous modernisation of legacy systems—supported by Model-Driven Engineering and DevOps—can replace costly full-system overhauls with a more predictable, iterative approach.
Automation has always followed three classic phases. In the beginning, humans build the example output. Next, machines are designed to reproduce that output at scale. Finally, the machines are programmed to alert humans when quality is breached. This approach—known as Jidoka or “automation with a human touch”—ensures people remain central to quality control, even as processes are automated.
Take a simple sentence: “The chicken is ready to eat.” Without context, the meaning is ambiguous. Is the chicken cooked and ready to be eaten? Or is the chicken itself hungry? For AI, these uncertainties can significantly raise the risk of errors.
To manage this risk, AI can be applied across all three phases of automation—always with humans guiding the process. In Phase 1, AI can assist in generating outputs, while humans maintain final quality control. In Phase 2, AI can help build the “arm” of automation, increasing speed while ensuring consistency. In Phase 3, an independent AI can act as a risk-aware quality assurance agent, monitoring and verifying outputs.
We’ve already seen this in practice. At WorkingMouse, we tested how AI could improve code quality. Our Local AI generated the first draft of code. ChatGPT then reviewed it, suggested improvements, and those suggestions were fed back into Local AI. The revised code was reviewed again, and this time ChatGPT confirmed the result was 100% correct.
This is Jidoka reimagined for the age of AI—humans and machines working together to deliver high-quality outcomes. With AI assisting at every level, and humans setting the standard for quality, automation becomes more reliable, adaptable, and effective.
