Generative AI Language Models (LLMs) possess an incredible ability to comprehend textual content and extract valuable information, making them a crucial asset for dealing with large volumes of data. Leveraging LLMs like ChatGPT enables businesses to efficiently analyze texts and uncover valuable insights that might have remained hidden otherwise. Let’s explore some practical examples of how businesses can use LLMs to enhance their resilience from both the demand and supply sides:
Example 1 — Reviewing Customer Feedback and Identifying Actions:
Objective: The goal is to review customer feedback records and identify potential actions and their respective owners. By inputting a customer feedback record into the LLM API, the output will be structured in a usable format, containing relevant information on actions and action owners. In the below prompt a sample customer feedback is provided for illustration (Courtesy: Kaggle customer feedback datasets) and the output is formatted into SQL statements to insert the records in a table for further action.
Prompt:
Given the following customer feedback within < >:
<
“Worthless, except as a regular echo and a poor excuse for video chat. I love my echo devices, bathroom, pool, kitchen, other places where I may need hands-free, voice-activated music and info. My wife bought me the ‘newest, hottest’ thing. I was skeptical but then thought I would use it to help on a project. Me ‘Alexa find videos on f450 drone’ Alexa ‘YouTube is not available’. Amazon won’t directly sell chrome products, YouTube won’t play on the echo show. Further testing shows the video call is more limited than iPhone or Android apps for video. So the most useful thing now is the same voice functions that my echos and dots perform. Unless I want to make all of my video calls and check the weather from a device I can move no more than 3 feet from an outlet.”
>
Identify the sentiment, extract useful information such as the product, and suggest actions that need to be taken. Create SQL statements to insert the extracted information into the feedbackAction table, including attributes like Product, Date, Sentiment, Action, and ActionOwner. For each individual action, a new record needs to be inserted, with potential action owners being the ‘Product Design’ team, ‘Logistics’ team, ‘Quality Audit’ team, etc.
Example 2 — Extracting Risk Information from News Articles:
Objective: In this example, we aim to extract useful information from news articles to identify potential supply chain risks. By extracting relevant articles using APIs like gdelt (https://blog.gdeltproject.org/gdelt-doc-2-0-api-debuts/) in a program workflow, we can utilize the ChatGPT API to extract risk details and sentiment for each business mentioned in the provided article.
Prompt:
Given the following article within < >:
< Insert article here>
Extract useful information and determine the sentiment of the article. The response should be prepared in JSON format as shown below for every company mentioned in the article along with other details such as company location, event and sentiment:
{
“company”: Company,
“location”: Location,
“event”: Event,
“sentiment”: Sentiment
}
Here is a link to a news article to try out as an example:
Sample output based on the above article content:
Try these example prompts and witness the immense value LLM APIs can generate for your organization. These examples are indicative, and you can adapt them to extract relevant business-specific information that aligns with your workflow.
Conclusion:
Generative AI LLMs like ChatGPT hold tremendous potential for businesses, enabling the extraction of valuable insights accurately, cost-effectively, and in a time-efficient manner. By incorporating LLMs into various workflows, businesses can enhance their resilience by making data-driven decisions and uncovering critical information hidden within vast amounts of textual data.
