+AI Insights

Contents

Overview

This article is about creating your own custom AI model using training data to best meet your needs. The purpose of a +AI Insight model can depend on the scope. A model within the scope of a UTA will appear in the specific UTA it was created for.

A UTA level model can perform data analysis and reporting across multiple records within a single UTA. This will give you the ability to quickly query your dataset using natural language instead of building a list view filter or report. For example, you could ask +AI Insight a question like “Can you list all the applications that are still in the ‘Draft’ status?” or “What are the award funding amounts for 2020, 2021, and 2023 respectively?”. Using the answers provided by +AI Insights can help you in several areas, such as aggregation, summarization, and prediction. It also provides the added benefit of using your own custom training set so the AI model can learn to incorporate knowledge specific to your process within the system.

A model with the scope of an entire instance will appear within your system header. Instance level +AI Insight models contain uploaded training data. Instance level +AI Insight models are ideal for uploading operational policies and procedures, governance policies and protocols, system usage guidelines and program specific guidelines.

In this article, you will learn how to set up both types of custom AI models.

Note: You will need an OpenAI license to use AI features on SmartSimple. You must be a Global Administrator to enable this feature. Contact your account manager or sales@smartsimple.com for further information on billing and implementation. If you have an AI implementation with Microsoft Azure, be aware that some SmartSimple +AI features will not be available to you, such as +AI Insights.

Configuration

Configuring +AI Insight for UTAs

To configure a +AI Insight model for a specific UTA, follow these steps:

1. Navigate to Global Settings > +AI tab > +AI Insights.
2. Click the “New Insight Model” button (plus sign).
3. Under the Scope field, select “UTA”. With this configuration, the model data is based on Level 1 records of a specific status and type and will only be available to a single UTA.
4. Under the Name field, enter a descriptive name.
5. Under the AI Instructions field, tell the AI what you want it to do with the UTA data (Example: "Analyze grant applications and provide aggregation data").
6. Under Conversation Starters, enter any common prompts a user may wish to ask the +AI Insight model (Example: "What is the total of requested funds for applications currently in the approved status?"). Separate each conversation starter with a line break. These conversation starters will be displayed on the chat interface for users to select.
7. Under Tracking Application, select the desired UTA. If you wish to filter the kinds of records that will be used by the AI as training data, you can do so by using the Type and Status fields.
8. Set role permissions as desired. .
9. Click Upload Training Data to update the AI’s model with the desired Level 1 records. When uploading a JSON file, be sure to enclose name/value pairs in double quotation marks ( " ).
10. Click Save.

Note: To ensure the AI is using the most recent dataset, be sure to periodically press the Upload Training Data button to update the AI with the latest information from the selected Tracking Application.

Once the above configuration has been set up, users can access +AI Insight by navigating to the appropriate UTA and clicking the +AI Insights button in the action bar. If multiple +AI Insights have been configured for the UTA, the user can select the desired model from the dropdown to launch the chat window.

Configuring +AI Insight for Your Instance

To configure a +AI Insight model available to your entire instance, follow these steps:

1. Navigate to Global Settings > +AI tab > +AI Insights.
2. Click the “New Insight Model” button (plus sign)
3. Under the Scope field, select “Instance”. With this configuration, the model data is based on uploaded training data. Instance-wide +AI Insight models can be accessed through the global header by users.
4. Under the Name field, enter a descriptive name such as “Organizational Policies and Procedures”.
5. Under the AI Instructions field, tell the AI what you want it to do with the uploaded training data (Example: "Analyze the uploaded data and provide answers to user queries").
6. Under Conversation Starters, enter any common prompts a user may wish to ask the +AI Insight model (Example: "Outline the process for contesting a decision"). Separate each conversation starter with a line break. These conversation starters will be displayed on the chat interface for users to select.
7. Set role permissions as desired.
8. Next, go to the “Training Files” tab and upload all the relevant documents the AI will need in order to answer in the desired manner. When uploading a JSON file, be sure to enclose name/value pairs in double quotation marks ( " ).
9. Click Save.

Note: In regards to uploading training files, if uploading a PDF file, avoid using tables and make sure all the text is selectable for the best results. Convert .doc files to .docx before uploading.

Once the above configuration has been set up, users can access the instance-wide +AI Insight by clicking the +AI Insights button in the global header. This will launch the desired model in a new chat window.

Use Cases

Reporting with +AI Insights

Instead of setting up a report, here are some examples of questions you could ask +AI Insight:

• Make a list of the top ten funding recipients by amount
• Which applications with a status of “Review” have been submitted within the last 30 days?
• How many applications had their status set to “Declined” this month? Make a 200 word summary of why they were declined.
• Using all the applications that have their status set to “Approved” this month, write a summary of why they were approved.

Compare/Contrast Application Submissions

As an example, let’s use the +AI Insights to help reviewers compare and contrast application summaries on a grants manager (or similar submission tracking application). Unlike the Work with +AI feature (ideally suited for external users) that can help you on the individual record page, +AI Insights can aggregate, summarize, and review data across multiple records.

If we have multiple records in our grants manager with each record containing a text summary of their video applications, we can ask +AI Insights to compare and contrast these application summaries to help us in our decision-making process.

1. Go to the UTA list view.
2. Select the +AI Insights button in the top action bar to select and launch your desired +AI Insight model.
3. In the chat window, ask +AI Insight compare and contrast the text transcripts of two applications and make a recommendation based on how well the text transcript meets program objectives.

Note: If your submission process includes media files, you can use other AI features to automatically generate summaries of the media contents as soon as they are uploaded. To find out more, see +AI Vision and +AI Transcription.

Surface +AI Insights on Portal Pages

+AI Insight models can also be surfaced in portal sections to allow users to interact with a specific +AI Insight model by presenting a button, shortcut, or by with a direct link within the portal interface. To configure this feature as a button, follow these steps:

1. Go to Main Menu > Users tab > Portals > Edit the desired portal > Create or edit a portal section with buttons such as a banner type section.
2. Under the Buttons field, add a new button and give the button a descriptive Caption (Example: "Chat with +AI Insight")
3. Click the URL Lookup button (binoculars icon)
4. In the URL Lookup window, select "+AI Insights" as the object type. Select the desired scope (UTA / Instance / Global)
5. If you picked UTA select the desire tracking application and specific +AI Insight model you want to surface.
6. Click Apply to save the selection

If the user possesses the necessary permissions to access the +AI Insight model, clicking the configured button in their portal will open a new or current window dedicated to the selected +AI Insight model, enabling the user to enter prompts, receive aggregation, and generate summaries or prediction information.

Alternatively, you may surface the +AI Insight model as a portal shortcut, or direct link similarly as described above.