Leveraging AI for efficient waste identification processes

Identifying and eliminating waste has long been a cornerstone of operational improvement. Traditional methods of waste identification have proven invaluable but often come with significant time and resource costs.

In this blog post, I break down a real-life waste elimination business case – IT service incident root cause analysis – to showcase how a half-century-old waste identification analysis can be revitalised by a small investment in AI.

Mikko Valtonen is an Aalto University graduate, a quality black belt in processes, and helps customers at Vuono Group to set improvement targets, execute change, and measure success using the best AI tools on the market. At Vuono Group, we optimise processes through data and software to drive business value.

What is waste, and why does it matter?

The Honda Accord became the best-selling car in the USA in 1989. In the 1990s, the deteriorating profits of the US auto industry, the emergence of the Rust Belt, and the success of Japan led to a desire to learn from Japanese companies. One of the most popular and powerful teachings at the time was Kaizen continuous improvement. Kaizen is a common-sense method based on classical empirical science.

Kaizen, Source: Wikipedia 

The Kaizen method manifests the importance of identifying and eliminating waste. Below is a list of waste types collected from multiple authors:

• Internal quality issues: The Six Sigma process school emphasises that only 3.4 out of 1 million business events can violate quality standards.

• Lost time waiting for input: Agile manufacturing, Agile development, and DevOps methods all emphasise this point. One of the best ways to reduce input delay is cross train “full stack competent” resources that require less input to complete their work. This can be seen to include wasted time on customers waiting for delivery.

• Lost time due to unrealised acceleration potential: Just-in-time and lean manufacturing principles emphasise the importance of short lead times. This is also highlighted in the business process re-engineering method.

• Dumb requirements: Requirements from a "smart person" are often the most dangerous since you might not question them enough. Eliminating requirements was the theme of the father of Business process re-engineering, M. Hammer, in his book “Reengineering Work: Don't Automate, Obliterate”, 1990.

• Unnecessary part or process: Each requirement or constraint must be accountable to a person, not to a department, because you can ask that person about its relevance and purpose, rather than having a requirement that nobody owns and persists for years despite being redundant. Eliminating tasks was the theme of the father of Business process re-engineering, M. Hammer, in his book Reengineering Work: Don't Automate, Obliterate", 1990.

• Simplify or optimise the design: One of the most common mistakes that smart engineers make is to try optimising something that should not exist. Instead, requirement and task deletion should be evaluated before trying to optimise.

• Wasted time on manual routine that could be automated: This is the leading thesis in all IT investments.

• Customer reclamations.

One of the biggest challenges of the Kaizen method is the cost of identifying waste. Identifying waste requires a team, which is an overhead cost. This team can become waste itself. Waste identification is an empirical science and is based on data.

Collecting evidence will require analysing the content of hundreds of thousands of quality documents, memos, incident resolution discussions, customer reclamation reports, problem tickets, change tickets, and other documents. Historically, this has been manual labour, and no computer tools could have processed unstructured documents.

AI tools for identifying waste

For the entire history of business improvement, studying these quality documents has been manual, slow, and expensive. Using AI for root cause analysis and complex classification has been studied since the 1950s, but the results were poor.

In February 2023, a ChatGPT moment removed this serious limitation. A startup called OpenAI showed that AI can be applied to complex transformations. As of May 2024, the smartest AI tools are listed below. Source: https://chat.lmsys.org/?leaderboard

These AI tools are impressive, but they still cannot perform most business tasks, and there are not many real successful AI use cases. The biggest reason why AI does not get the job done is the inherent random hallucination feature. Most tasks are sensitive to quality and reliability, and therefore, current AI agents are not suitable for them.

AI cannot (yet) perform your business tasks, but it can handle your transaction document analysis tasks. AI can handle all the work where counter mistakes compensate each other.

Case: IT service incident root cause analysis with AI

If you're reading this as an IT professional, I recommend starting by identifying waste in your own backyard. In IT, all the waste is documented quite well, and data is available in IT systems.

In this example, each ServiceNow incident represents a documented waste. In ITIL (Information Technology Infrastructure Library), an incident is defined as any unplanned interruption to an IT service or reduction in the quality of an IT service.

Incidents may be reported by users or detected by event monitoring tools. Each incident represents a disruption or potential disruption to the normal operation of a service and requires a response to restore normal service operation as quickly as possible.

Using AI to scan through each and every incident is a relatively small investment. We have completed the setup below in a one-week project:

Steps 1 and 2

Each ServiceNow incident is sent to a gold standard AI, such as Azure OpenAI API (or an alternative AI API like Claude 3).

Step 3

AI identifies and classifies incident root causes.

Step 4

These root causes are then stored in an Excel sheet.

Step 5

The root causes are further visualised in a pivot table (below). The visualisation reveals a typical 20/80 balance between waste driver root cause and waste.

The 20/80 balance between waste driver root cause and waste

Typically, incident content analysis reveals at least some easy-to-fix top waste drivers. Some of these low-hanging fruits are commonly related to master data quality. A typical actionable insight is that a small change in collaboration will lead to the elimination of frequently recurring root causes.

Conclusion

Integrating AI tools like Azure OpenAI API into waste identification processes represents a significant leap forward in business optimisation. By automating the analysis of incident reports and other quality documents, organisations can uncover and address inefficiencies more swiftly and effectively than ever before.

As demonstrated, even a modest investment in AI technology can yield substantial benefits, transforming waste management from a labour-intensive task into a streamlined, data-driven process.