IPTD-469
NJDEP Policies and Procedures (now ALL DEPnet) chatbot
Description
This project will employ an LLM to produce a custom chatbot that will be fed the content of DEP's internal policies and procedures to enable staff to ask anything about those P&Ps. If the associated internship and available resources permit it, the chatbot may be extended to encompass all DEPnet content.
Problem Statement
NJDEP Policies and Procedures are extensive and not always worded for easy general consumption. The wording of policies leans toward more legalistic expressions as they may be called upon within supervisor/employee disputes, especially the large number of Human Resource policies. Employees (especially new employees) with questions cannot readily find clear answers to situations that will be new to them.
Project Justification
This project is the focus of intern Manav Parikh and while satisfying his internship demands, it is also expected to advance DOIT's understanding and capacity with AI/LLMs and DEP content.
Also, the experience of new or potential future full-time employees (such as interns) is a focus and priority for leadership eager to make NJDEP a desirable destination for talent. This project would produce a product that demonstrates DEP is progressive with technology including technology deployed for benefit of the employees.
Also, the experience of new or potential future full-time employees (such as interns) is a focus and priority for leadership eager to make NJDEP a desirable destination for talent. This project would produce a product that demonstrates DEP is progressive with technology including technology deployed for benefit of the employees.
Estimated Transactions
None
Components
Large Language Model (LLM)
Target Rollout Date
None
Target Rollout Date Reason
None
Activity
Show:
After the support of two internship periods and now an hourly (Manav Parikh) the tool is evolving to a robust access to all DEPnet content the current beta-version is found here: AI_ChatBot_All_Conv – test17 current work is toward ensuring effective periodic updates from content changes across DEPnet. Updates are now being handled, but deletions need a different approach. We are also working to build in some useful feedback elements and already are showing the last editor to referenced content.