The challenge all manufacturers face is how to extract and preserve the priceless tacit knowledge of their workforce. Symbolic AI offers a solution by capturing specialized know-how and making it accessible, preventing decades of experience from being lost. XpertRule CEO and Founder Akeel Attar explains how.
Manufacturing’s greatest asset is its people, but when experienced employees leave, they take decades of hard-earned knowledge with them. With nearly half the workforce aged over 50 and many thousands set to retire in the next few years, this issue has never been more urgent.
While some may delay retirement or shift to part-time work, the challenge of transferring this invaluable expertise before it’s too late remains. Symbolic AI provides a powerful way to bridge this gap by capturing and preserving knowledge for future generations.
What is Symbolic AI?
While many people associate AI with tools like Generative AI and virtual assistants, there’s another type of AI – what I call ‘good old-fashioned AI’ – that has been evolving over the past 40 years: Symbolic AI.
Unlike Gen AI, which relies on vast amounts of data to identify patterns, make predictions and generate new content, Symbolic AI extracts human expertise and transforms it into structured, actionable frameworks. It is an acknowledgment that human know-how and data-driven insights are equally important when it comes to understanding and optimizing a process. This makes it ideal for industries like manufacturing, where human intuition and experience play a decisive role in keeping operations running smoothly.
In a factory environment, when something goes wrong with a machine, early warnings often appear in subtle ways – perhaps a strange noise or a spike in a reading. A seasoned operator knows exactly what corrective action to take based on these cues, but how do you capture that knowledge in a form that’s useful, accessible and repeatable?
Symbolic AI offers a structured approach to codify human expertise through symbols, heuristic rules (rules of thumb) and decision trees that show the relationship between outcomes and the factors influencing them. Think of it as a flow chart or troubleshooting guide that provides immediate, actionable advice – ‘If this happens, then do this.’ Visual aids can also be added, allowing operators to compare what they see with established norms.
Key Applications of Symbolic AI in Manufacturing:
- Automating Documented Knowledge
Symbolic AI can convert documented, rule-based knowledge such as compliance regulations, standard operating procedures and safety protocols into automated systems. For example, quality standards that are clearly defined but tedious to follow manually can be automated through Symbolic AI, ensuring consistency while alleviating the burden of manual checks.
- Capturing Tacit Expertise
The true power of Symbolic AI lies in its ability to capture and formalize deep expertise that comes from years of hands-on experience. Experts often act intuitively, unaware of how they make decisions, troubleshoot issues or assess situations. This makes tacit knowledge difficult to teach, time-consuming to document and rarely found in textbooks.
Building a Dynamic Knowledge Base
The biggest hurdle in creating a knowledge base is that people often struggle to articulate their intuitive knowledge. They may not consciously think of their expertise as a set of rules or decision trees, which limits how much they can volunteer in explicit terms. The challenge, then, is finding ways to help them articulate what they know. At XpertRule, we’ve refined a process – knowledge acquisition by rule induction – to overcome this limitation.
In our experience, experts typically find it easier to explain things using examples because that is how they tend to process information. Rather than ask people to manually create a full set of decision rules, which can be overwhelming, we prompt them to provide situations and explain how they would respond based on their instincts and experience. AI then analyzes these real-world scenarios and formalizes them into a logical decision-making framework such as a decision tree.
This process allows us to efficiently extract expertise, with experts focusing on sharing knowledge, and AI handling the heavy lifting of structuring that knowledge. The more experts feed their expertise into the system, the more comprehensive and nuanced the decision trees become.
This process can also help resolve conflicting methods across teams. A factory may have managers applying different approaches for assessing risk or addressing issues. By using Symbolic AI, these contradictions are surfaced, giving companies the opportunity to mediate and formalize a consistent set of organizational rules. In doing so, a company not only captures individuals’ expertise but also ensures that the knowledge base aligns with the company’s broader strategic objectives.
Once built, this knowledge base becomes a powerful tool for training apprentices and younger employees, guiding them through the decision-making process. Rather than simply being handed the answer, trainees learn through experience, allowing them to understand not only potential solutions but the rationale behind them. This guided method enables workers to gradually develop the intuition necessary to make informed decisions independently.
Filling Data Gaps with Expert Knowledge
Every factory will have processes where the consequences of failure are so severe – whether financial or safety-related – that failure simply cannot be allowed. If that is the case, how do you effectively train AI models to perform risk mitigation or analysis without access to related data? Once again, the answer lies in using Symbolic AI to capture the expertise of seasoned operators who’ve been doing this type of analysis, sometimes unconsciously, for decades.
In most cases, the best results come from combining data-driven models with Symbolic AI. This hybrid approach can offer greater insights, especially in data-scarce situations. For example, XpertRule worked with a complex machine manufacturer to automate remote service support for their Worldwide customer base.
The initial phase involved capturing the expertise of mechanical, control and process engineers and codifying the rules they used to evaluate applications. Using Symbolic AI, we were able to automate 50% of applications based on the engineers’ explicit rules.
When the company wanted to increase the support services, engineers found they couldn’t provide any additional rules. That’s where data came into play. By analysing past process data, we discovered patterns the engineers had applied unconsciously. We extracted these ‘hidden rules’, presented for validation by the engineering team and successfully increased the automation rate to 80%.
This meant the customer support team could handle more enquiries in less time without referring to the process team and could ensure despatch of the right resource on the occasions when an onsite visit was required.
Given the many challenges the industry faces, manufacturers can’t afford to lose critical expertise. By leveraging Symbolic AI, they can safeguard knowledge, ensure operational continuity and future-proof their business – no matter how their workforce evolves.
This article is part of our new series, Reality Check: What AI Really Means for Manufacturing, designed to inform, inspire and help you implement AI in your manufacturing operation.
Look out for our upcoming article where XpertRule’s Iain Crosley will round up key discussions from PharmaTech Integrates 2024. He’ll explore how AI is transforming drug development and manufacturing, including highlights from his session on practical applications of AI in Pharma and Life Sciences.
Oct 3, 2024 2:03:45 PM