Research Repository: Lowry

Background

As a researcher, often the sole researcher for my UX teams, I have long dealt with the ever-present problem of managing, storing, interpreting, and disseminating data from previous research projects and initiatives. Often times this can create a bottleneck or chokepoint for those wishing to access or ask questions of research. A solution that I began developing to address this problem was the formulation of an AI created agent coupled with a chatbot where colleagues could ask questions and get answers without relying on me or my availability. This agent, named Lowry in honor of The Giver by Lois Lowry, represent a medium through individuals can access the pains, emotions, needs, and findings of previous research.

Objective

Create an accessible format for employees to be able to ask questions of previous research on their own terms and explore the data with the help of a developed AI-agent.

    • All internal employees and colleagues

    • Research

    • Data Science

    • AI Platform Team

Something Interesting

The Giver, by Lois Lowry, is a tale of a dystopian society that has managed to shift all pain and strife to a sole holder of those emotions, know the Receiver of Memory. He is then responsible for shifting those memories to a new protege. Often a researcher can feel like the Receiver of Memory and has a responsibility to “give” those memories to others.

Method

This technical project is highlighted to present an innovative technique to solve an ancient problem for researchers.

Procedure: Inventory and Asset Collection

To commence the project, I wanted to establish a pilot program using select pieces of research data that had been gathered throughout the years. The most fruitful proving ground was longitudinal survey data from our annual Strategic Sales Survey. Nine years of data querying sellers on market and product positions were wrangled into a unified data set using aggregation and SQL.

It is noteworthy that data assets were mixed and varied. Formats included:

  • .txt

  • .xlsx

  • .ppt

  • .mp4

  • .html

Research Assets

Procedure: Development

Using OpenAI and Python scripting, I developed a test Agent (“Lowry”) that could look through the datasets and derive responses to questions via a chat-interface. Since this was a personal development project leveraging newly learned skills, I had to work through the intricacies of integrating with OpenAI and tuning Lowry through a series of test prompts and validation questions.

This section was the area where the most lessons learned occurred. I had not leveraged Python in this manner before and tuning an agent to not only accept certain data types but also respond accurately is not trivial.

Development

Results & Discussion

This project of professional and personal development was unfortunately quashed during the most recent company acquisition due to standards and prohibitions on using AI for work. However, the value was immediately recognized by several teams during our test phases. Prior to the acquisition, it was positioned as the preliminary pilot program for the use of AI for internal tools. Had it been seen through, I’m confident in its ability to not only help my colleagues understand user, client, and consumer perspective, but also help me and my team manage requests.