Using AI Sentiment to Prioritize Your Cases

Have you thought about using AI Sentiment Model to determine the priority of cases assigned to yourself? In Power Automate there is a prebuilt model that can perform a sentiment analysis on text. The model is called: Analyze positive or negative sentiment in text. This can be very handy action in determining if a certain text is either Positive, Negative or Neutral. In the world of business applications there are plenty of use cases where you can apply this, for example for analyzing incoming emails or posts on social media. I wanted to experiment my self with this functionality and came up with the following case:

On a helpdesk of a company that sells mobile phones and equipment they receive all kinds of requests, think about:

• questions from existing clients about products
• requests for repairs
• questions that could be answered by guides

There are lots of ways to determine what the right priority would be of handling these requests, that could be either First In First Out, an Entitlements, Rules around priority or perhaps other specific logic determined by the business. In this case we want to manage this on sentiment and in this specific case the Negative will have the highest priority.

In order to accomplish this, I have created a PowerAutomate Flow which will analyze the Case Description field as soon as the Case is created. You need to provide the case description and the language (in our case (en-us) to the model in order to determine the sentiment.

PowerAutomate Flow using AI Sentiment connector

In the Update actions, we will set the Sentiment as well as the score. The score is a value in the range of 0 to 1. Closer to the 1 indicates greater confidence that the determined sentiment is more accurate.

In order to create a view and being able to sort on the emoji (sentiment) I have created a calculated field that would add a sorting value to the record.

If you would now create a view that would show Cases based on Sentiment and then Sentiment Score, the ones one top of the list are the most Negative ones and should be picked up first.

I know that this is probably not the ideal situation for determining priority of cases, but I hope you found the article valuable.

Before you implement this to your own business, or similar logic, it would be good to check if your language (the text you want to analyze) is supported by the model.