We tried to assess the potential uses that a manufacturing organisation could make of chat-enabled generative AI technology and assessed the potential impacts of this technology for the industry by analysing the very answers an open tool has given us.
We questioned none other than ChatGPT, which stands for Chat Generative Pre-trained Transformer. ChatGPT is based on OpenAI’s GPT-3.5 series, which can perform a variety of natural language processes*.
With the professional version available, assuming that outage and availability problems are now past, does it make sense for a manufacturing organisation to fork out the few dollars of monthly subscription? Let’s find it out.
When we asked ChatGPT how it can be useful in manufacturing processes, the response listed the following areas:
“Quality control: ChatGPT can be used to analyse written reports and data from quality control inspections, identifying patterns and trends in product defects.
Predictive maintenance: ChatGPT can be used to analyse sensor data and logs from manufacturing equipment, identifying patterns and trends that indicate when maintenance is needed.
Document summarisation: ChatGPT can be used to automatically summarise large amounts of technical documents, such as manuals and procedures, making it easier for employees to find the information they need.
Natural Language Processing: ChatGPT can be used to process natural language input from users, such as voice commands, and translate them into actionable instructions for manufacturing robots or machines.
Chatbot: ChatGPT can be used to create a chatbot for customer service, handling customer inquiries and complaints.
Language Translation: ChatGPT can be used for language translation for multi-lingual communication with customers and suppliers.
These are just a few examples of how ChatGPT can be used in manufacturing. The specific application will depend on the needs of the manufacturing company and how they want to utilise the technology.“
We then proceeded in investigating some more details for each of the answers. The follow up question was “How can ChatGPT be used in quality management in manufacturing?”
We are now providing the actual answers, with a brief comment and our verdict.
“ChatGPT can be used in quality management in manufacturing in several ways. One way is by analysing data from quality control inspections to identify patterns and potential issues. The model can be trained on historical data to predict when quality issues are likely to occur, and can also be used to generate reports and alerts to notify quality control personnel.”
Any modern Quality Management System (QMS) application can already be augmented with an embedded AI analytics module to achieve this. This is very useful to help companies figure out complex root cause structures, and in general, investigate their “known” and “unknown” unknowns. In our case, it is unclear on which data Chat GPT is going to rely. So, at the present time, we think that ChatGPT is overselling a little bit here.
“Additionally, ChatGPT can be used to generate automated test plans and reports to increase the efficiency of quality control inspections. ChatGPT can also be used to generate automated documentation for compliance with regulations and industry standards. This can include generating standard operating procedures, work instructions, and other documents to ensure that quality management processes are followed consistently.”
This is quite interesting. This can help a lot with the simplification of clerical work and create multiple reports for different audiences very easily and at scale. But for the real business impacts to materialise, a company should first do some preparatory work and carefully define business rules, templates, formats, and audience requirements.
“Another way ChatGPT can be used in quality management is by analysing customer feedback and complaints, and identifying patterns and common issues. This information can be used to improve the quality of products, and to prevent similar issues from occurring in the future.”
The good old “feedback loop” story: understanding user habits and recurrent issues to feed back to the quality department with relevant information. But in the manufacturing industry, the problem for most companies is how to bypass distributors and access timely and good quality market information directly to feed the analysis. If ChatGPT only relies on publicly available information and does not tap-in any market intelligence database, everything it would talk about is things a company should already know about. So, as an add-on to a market intelligence subscription, it could become an intelligent way to consume the data, otherwise, mostly of little use.
“Overall, ChatGPT can be used to improve the efficiency and effectiveness of quality management processes in manufacturing, and help to identify and prevent quality issues before they occur.”
Here it seems a definitive oversell.
Then we followed up with the next use case: predictive maintenance and this is the answer.
“Analysing large amounts of data from manufacturing processes and equipment to predict when maintenance will be needed. This can include data from sensors on equipment, as well as data on past maintenance and repair history.
The model can also be trained on historical data to identify patterns and make predictions about when equipment is likely to fail.
Additionally, ChatGPT can be used to generate reports and alerts to notify maintenance personnel when maintenance is required, and can also be used to optimize maintenance schedules to minimize downtime and improve overall equipment efficiency.”
We are not doing a point-by-point analysis here, but in general, we are seeing the same pattern as the use case before. Overall, ChatGPT can be used to improve the efficiency and effectiveness of document management processes in manufacturing and help to identify and extract important information from documents quickly and accurately.
Good for clerical work simplification, redundant for other functionalities, but definitely overselling itself sometimes.
Now, let’s analyse the other points that ChatGPT has brough forward: Document summarisation, Natural Language Processing, Chatbot, Language Translation.
A quick glance at the detailed answers highlighted that here ChatGPT was not providing additional business use cases, but rather outlining some basic principles of how it works. A little bit like if we asked an intelligent washer-dryer machine what it does, and the reply was: “Washing clothes, drying clothes, having a spinning drum inside, being heavy, consuming electricity, etc.”
Any Missing Areas?
Now we also thought about some obvious “missing areas” in its self-assessment: for example, idea generation. Using the generative text functionalities as a way to explore concepts, visions, user needs, like a “crowdsourcing” tool that prompts ideas that may just not come to mind to a decision maker considering several options.
Another obvious miss was code programming. Our research has constantly highlighted a critical need for digital skills in factories and the importance of low/no-code applications to address this (we have been talking about the notion of “buyilders” — a hybrid between “build” or “buy” software users), enabling workers on the shop floor in general act as/programmed as interactive tool/personal assistant, digital work instructions, etc…
As personnel with some knowledge of coding become more of a frequent sight in plants, having a conversational tool for programming can easily lower the threshold for digital skills in disciplines such as data analysis, machine connectivity, and operational improvement.
What Are the Risks that a Manufacturing Company May Incur While Using this Type of Solution in Their Business Practices?
First, the risk of losing the grip on information quality. Over time, managers can lose grip on the practical aspects of information they have to evaluate (where it comes from, how it has been created, are there inherent biases in it? etc.), with an increased risk of “garbage in/out” effect (a side note: with examples of answers such as “gnothi seauton”, or “42!” we can think of plenty of “oracles” providing unintelligible answers to poorly structured questions).
It is true that sloppy interns will get wrong information as well, but the system has no way to check the quality of information, at least not yet. Eventually, there is a risk of having supervisors that have never done the job in the first place because the step is not done anymore.
Also, these tools automate a lot of “entry-level”, clerical, repetitive transactional tasks (filling repetitive reports, data entry, data check, etc.). This is not bad per se, but it could be an issue in organisations where the “learning by doing” career growth is heavily based on those menial activities.
The risk is that there could be no available route for juniors to learn and grow in these organisations.
Hampering radical innovation reduces the questioning of the status quo. The widespread use of generative tools will also force companies to reinvent excellence.
In its answers, ChatGPT tends to converge to a median level, the “common knowledge”. So not only can this lead to a risk of preventing “disruptive discovery” if not used properly, it can also amplify and reinforce organisational biases, favouring routine and procedural consistency over innovation. Removing the transactional costs of procedures (i.e., clerical work is boring and expensive) can make organisational streamlining less of a need.
Making the fulfilment of useless steps automated and cheap may lead to the accumulation of inefficiencies. When nobody has the ownership and burden to execute these tasks, nobody questions the procedure, and the risk of procedural errors or bottlenecks arising from convoluted processes is never addressed.
The advice is to take this technology with a grain of salt. Companies have to look at it with an eye for process transformation. Automating something manual is not a benefit by itself, it is a benefit when it simplifies the organisational structure, saves time, and increases the focus of key decision makers to more relevant tasks that create real business value.