About how we asked ChatGPT to help with the FMEA
Everyone is talking about ChatGPT and it’s impossible to ignore the topic. Many people wonder about the importance of this innovation. Some conservatively dismiss the eggs potential by locking themselves in a bubble of temporary comfort. Browsing the Internet, we’ve came across an interesting comparison. Although, unfortunately, we didn’t note the author, the quote was: “ChatGPT will not eliminate your job, just as Excel didn’t make the accountants disappear. Only those who haven’t learned to use it disappeared.” Following this line of thought, it’s worth learning to use this amazing new tool.
We decided to see how ChatGPT can help us do something really worthwhile, something that is useful in our work. At Softgent, we use FMEA, or Failure Modes and Effects Analysis, as a valuable risk assessment tool, helpful in identifying and preventing potential failures in a certain process. The tool, while useful, is not one of the easiest and pleasant ones. The output document can be very complicated and can contain many lines, especially for complex processes.
We’ve decided to see if ChatGPT would help us prepare an FMEA. The choise was software development process. First question to ChatGPT was:
- Provide potential failure modes for software development process for the purpose of FMEA
-
Sure, here are some potential failure modes for the software development process that could be considered for the purpose of FMEA:
1. Insufficient requirements gathering: The team may not fully understand the user's needs or requirements, leading to a product that does not meet user needs or expectations. (...)
Right. First potential failure mode defined: Insufficient requirements gathering. We all know this one, right? Let’s continue:
- List potential effects of failure for insufficient requirements gathering for the purpose of FMEA
- Product does not meet user needs: If the requirements are not properly understood, the resulting product may not meet the user's needs or expectations, leading to dissatisfaction and potential loss of customers. (...)
Let’s go for the easy way. The first suggestion creates the first line in the FMEA. So we have a product that doesn’t meet the needs of the user.
- Provide potential cause(s)/ mechanism(s) of failure for product that does not meet user needs for the purpose of FMEA
- Lack of user involvement: If the product team does not involve users in the requirements gathering process, they may not fully understand user needs, leading to a product that does not meet their needs. (...)
Great we have a lack of user involvement. It’s time to define prevention.
- For the purpose of FMEA please define Current Controls (Prevention) for Lack of user involvement
- User research: Conducting user research to better understand user needs and preferences, and using this information to inform product development decisions. (...)
It seems quite right – but how do we detect it?
- For the purpose of FMEA please define Current Controls (Detection) for conducting user research to better understand user needs and preferences
- User feedback mechanisms: Providing mechanisms for users to provide feedback on the product, such as surveys, user testing, or focus groups. User analytics: Collecting and analyzing user data, such as website analytics or product usage data, to identify user behaviors and preferences. (...)
Now let’s put it together and see what came out of it all:
| Potential Failure Mode | Insufficient requirements gathering |
| Potential Effect(s) of Failure | Product does not meet the needs of the user |
| Potential Cause(s)/ Mechanism(s) of Failure | Lack of user involvement |
| Current Controls (Prevention) | User research |
| Current Controls (Detection) | User feedback User analytics |
According to ChatGPT, the software development process may not work because we have not gathered the requirements for the project sufficiently. It may result in us producing software that does not meet users’ expectations. A potential reason for this could be not gathering enough feedback from potential users before we’ve started our work. Of course, the best way to prevent this is to study user preferences. To this end, ChatGPT suggests implementing mechanisms for collecting feedback from users, using analytical tools for tracking preferences, etc.
Apparently, ChatGPT hasn’t discovered anything new, but one has to be admitted – it makes sensible suggestions. It doesn’t relieve us from thinking, but it is a time saver for sure. We definitely see a lot of potential in it. We leave the rest to your evaluation…