A3 Reports

With the advent of AI I am very interested in how problem solving will evolve. AI coaching and review of problem-solving content is already possible. AI assistance in drafting reports is also becoming common. That does not mean AI authoring the report. It means AI assistance in writing the report and in using HTML, SVG, and other tools to render displays of information. This was always possible with high-end graphics packages and 3D CAD systems. Now it is possible with LLMs and related tools.

From time to time I will post A3 reports here that experiment with AI tools for depiction. Each one works from documented source material. The machine-shop scrap case is restored from the original in A3 Thinking. The Ohno 5-Why case is a historical recreation drawn from several sources, including Ohno's book and Toyota personnel. The Apollo 13 and Challenger cases are built from the official NASA investigation records. The roundness case uses synthetic example data. Where a detail is reconstructed for teaching purposes, the report says so. The Apollo 13 introduction page explains the first experiment in more detail.

Some predictions

AI will increasingly be used to generate A3 report content. Some of it will be poor: one-shot images produced by distillation, suitable for social media and not much else. Other changes will be for the better.

• Format. We will see more use of the vertical, scrolling format alongside the traditional horizontal A3 sheet.
• Quantitative displays. There will be better ways to create visual displays of quantitative information.
• Depth on demand. Hyperlinks let a report embed supporting information that the reader clicks only when more depth is needed.
• Interactive data. The reader hovers a Pareto bar or a point on a control chart and sees the underlying numbers, or filters the data by shift, machine, or date.
• Animation and simulation. Process flow, spaghetti diagrams, and changeover sequences shown in motion rather than as arrows on a drawing.
• 3D depictions. Parts, fixtures, tooling, and layouts can be shown as 3D objects the reader can rotate and inspect, rather than a flat sketch or photo.
• Video. AI video editing is already possible. Embedding video in reports, for example to show details at the genba, will eventually become regular practice.
• AI-checkable logic. If the report is structured data as well as prose, AI review can check whether the problem statement, the cause chain, and the countermeasures actually connect.
• Instant translation. The same report in Japanese and English for plants on both sides. AI translation is already good enough for this.