Let me show you how I use a fine-tuned GPT model to enhance the onboarding process for new team members and simplify knowledge base interaction in my team.
Imagine a scenario where you have a large knowledge base for your project, and a big team that interacts with it on a daily basis. Team members often need to check how existing processes should work or have questions about the project and organization in general. To automate and increase the efficiency of this interaction, I developed a chatbot based on the OpenAI GPT model, fine-tuned to address specific business needs.
The chatbot saves a significant amount of time for colleagues and boosts the overall efficiency of the team. However, due to confidentiality reasons, I cannot share the knowledge base used in my company. Therefore, I created a separate knowledge base in a secure cloud.
Let’s consider a specific example. Imagine a page in the knowledge base called ‘Ways of Working’, which includes a section on user story workflow. A user story is a brief and simple description of a feature or functionality from a user’s perspective. User stories are used in agile software development to capture requirements and prioritize tasks based on user needs.
A user story workflow is a sequence of steps or stages that outline how a user story is completed from start to finish. Throughout this workflow, a user story can move back and forth between stages if adjustments are needed or if issues arise. Jira allows teams to customize workflows to merge their specific development processes and ensure efficient collaboration and tracking of tasks.
To have a clear understanding of who can move a user story from one status to another and under what circumstances, it is essential to have a well-defined workflow. This is where the chatbot comes in. By fine-tuning the existing GPT model developed by OpenAI, you can integrate the chatbot with your corporate messenger, such as Slack.
In this example, let’s assume I am a regular employee who wants to get more information about the current user story workflow. I can ask the chatbot simple questions like the order in which user stories are moved from one state to another. The chatbot provides the answer that user stories are moved from the backlog to the ready for development state, then to the in development state, followed by the ready for test state, the in test state, and finally, the ready for acceptance state. This aligns with what we have in our knowledge base.
As a new employee, I don’t need to disturb my team members to get answers to my questions. Everyone in the team is happy, and the workflow remains smooth.
Another important question is what conditions need to be met for a user story to be moved into the accepted state. This is crucial for the team to stay aligned and ensure that user stories are considered completed. The chatbot confirms that a user story can be moved into the accepted state when it meets the definition of done, which includes implementing requirements, passing key testing, presenting to the product owner, and having no critical or major severity defects.
Lastly, let’s address the question of who owns a user story when it is in the backlog state. The chatbot responds that, according to our processes and knowledge base, the ownership belongs to the product owner. They are responsible for defining the user story, describing its scenarios, and ensuring readiness for development.
This example demonstrates how you can use a fine-tuned GPT model in your everyday activities and interactions with the knowledge base. By integrating it with your corporate messenger, such as Slack, you can provide quick and accurate answers to common questions, saving time and improving overall team efficiency.
I hope you found this article helpful. If you have any questions or comments, feel free to leave them below. Don’t forget to follow our channel for more interesting videos. Have a great day!
