Quick view on the homepage
A summary of the Prism analysis (of Comments, currently) appears on the homepage. For a more detailed view, you will need to go to the Comments report.
How to Use the AI to Analyse Comments
Step 1: Access the Comments Reports
Start by navigating to the Comments reports section in your dashboard. From there, select the Question and the Filtered View that aligns with the data you want to analyse. For example:
Step 2: Review categories and summaries
Once you've made your selection, the results will display various categories and an overview summary.
Each category will feature a roundel, showing the percentage of comments within your filtered view that relate to that specific category. For instance, if the roundel indicates 10% for "Systems and Process" under "Finance & Audit," this means 10% of the comments in that view are focused on Systems and Process.
At the top of the results, you'll find an overview summary of all comments within the selected or filtered view. Below that, each category has a text summary of the comments, with sentiment colour-coded in green or red (if sentiment analysis is enabled).
Note:📍The results from one summary to another, as well as the recommended actions, may vary because Prism is not deterministic. This means that it generates responses based on patterns and probabilities, rather than following a fixed set of rules or outcomes.
As a result, even with the same input, Prism might produce different outputs each time, reflecting slight variations in how it interprets and processes the data.
This inherent flexibility allows the AI to provide diverse perspectives but also means that results can differ across sessions or prompts.
Step 3: Help with what actions you might want to take
You can create an suggestion for an action by clicking on this icon and it will open the chat window:
Clicking yes, opens up the action plan:
Step 4: Interrogate your data with chat
You can delve deeper into your data using the chat feature. This allows you to ask specific questions about your data, understand what's been said and by whom, and even pull out verbatim quotes. You can request tables, percentage splits, and further analysis to gain deeper insights.
💡Keep in mind that while the AI is powerful, it can occasionally make errors. To ensure accuracy, try rephrasing your questions and cross-checking the responses you receive.
Keep in mind that while the AI is powerful, it can occasionally make errors. To ensure accuracy, try rephrasing your questions and cross-checking the responses you receive.
By following these steps, you can effectively analyse and gain actionable insights from your comments.
About the Chat and Summaries
The Prism chat feature and comment summaries in your dashboard are generated using a People Insight version of advanced Generative AI (Gen-AI) technology from OpenAI's enterprise services. This AI is designed to help you quickly analyse and gain insights from your data, making it easier to understand complex feedback and take informed actions.
For more information on how, our data sub processor, OpenAI, handles data security, privacy, and compliance, you can visit the OpenAI Trust Center.
About the Categorisation and Sentiment Analysis
The categorisation of comments within your dashboard is powered by our own custom model, which we train and update using Microsoft Language Studio. This allows us to tailor the categorisation specifically to your needs, ensuring more accurate and relevant results. Sentiment analysis and Personally Identifiable Information (PII) redaction are also handled through Microsoft Language Studio services. Microsoft acts as a sub-processor for these tasks and operates within the UK, ensuring that data processing aligns with regional compliance standards.
Data Anonymity and Privacy
It's important to note that all data sent to these AI services is anonymised. This means that the information processed by our sub processor services is not attributable to any specific company or individual, ensuring your data remains secure and confidential throughout the analysis process.