Train ChatGPT on your client's Maximo knowledge base
A sneak peek into a Maximo AI chatbot, built on ChatGPT, which is custom-trained on the client's Maximo knowledge base
By the means of this blog, my aim is to show that we can have a chatbot that one may use as the single source of information related to the Maximo application of any business.
If a Maximo end user clicks on the icon at the bottom-right corner of the login screen, they can access the chatbot and use it for various purpose. To list a few:
Information on deployed Enhancement and Changes in Production
Way of working e.g. understanding work order creation to completion
Known issues and resolutions, etc.
This ChatBot can be trained by Open AI text-embedding models using the training documents that are initially provided. It has to be updated after each production go-live.
For example: You can find out what modules have been updated, what are the new features, or what changes have been made during the month/year.
You can also get into the details of the changes made.
The output is very specific and limited to the information provided in the training documents to ensure creditability and accuracy. The response is not influenced by different information on the internet and the main objective of educating on client-specific business flow is fulfilled. The training documents are stored locally and not on OpenAI servers. When someone will run the code for the first time, the chatbot will train itself and create an index.json file locally. When the training data is updated, they will have to re-create this file. The quality and efficiency of the response depends greatly on the training data.
If the end-user gets incorrect or incomplete response, they have an option to flag/report it. The developer can review the feedback and fix it accordingly.
There is also an option to go for fine-tuning models in order to improve the quality of response, and make the system cost effective in the long run.
Lets keep sharing new ways of optimizing Maximo and its business processes using these AI tools.
What do you think of this idea ? Leave your comments below. Thanks and please subscribe to the blog.
Good idea!. Did you already implement in any project?