Spotlight AI - Stand Alone Guide
- Dave Ruel
þSpotlight AI Stand-Alone is designed to allow you to import review-ready productions, and help you craft targeted, review-focused questions and export data intelligence to be used at the review level. At its core, Spotlight AI is built upon three important tenants, security, transparency, and purposeful use.
Disclaimer: Spotlight AI is intended as a tool to assist teams with data relevance assessment and should not be solely relied upon for making legal decisions or determinations. Hanzo does not warrant that the tool will be error free, and users should exercise their own professional judgment when deciding on case strategy.
Security
Spotlight AI Stand-Alone prioritizes data security by separating case description information and question generation from imported matter data. Case descriptions and questions are generated via a secure connection between Spotlight AI Stand-Alone and IBM Watsonx. All communication between Hanzo and IBM is encrypted; no user-created data ever touches the IBM environment. Additionally, all case descriptions and questions are ephemeral in nature and are not saved within the IBM tenant. The case description and questions are solely stored within your Hanzo tenant (GCP) environment.
Imported matter data is stored in your Hanzo Google Cloud Platform tenant, again isolated and shielded from external access. Spotlight data is rigorously encrypted. This multi-layered approach guarantees the highest level of data security and privacy for your interactions with Spotlight AI.
When to use Spotlight AI
Spotlight AI Stand-Alone is intended to help elevate the most important and relevant information to the very top of your imported data stack. Spotlight can be used on typical litigation matters or internal investigations such as IP, theft, disputes, etc. It has been built to be flexible with auto-generated questions or manual mode whereby you can input your own questions about your matter.
Matter Creation
Log into your system
Create a new matter by selecting the “Add New Matter” button in the upper right corner of the interface.
Enter a Matter Name and Matter Description. Select Load File from within the Data Sources list.
NOTE: The From Date, To date, Admin Access Group and Reader Access Groups are disabled within Spotlight AI Stand-Alone
Select Submit to create your matter
You will be directed to your newly created matter details page.
Load File Import
From your Matter Details page, select the Source Management tab at the top of the page. This tab will allow you to add load file sources to your matter as well as display any previously imported load files.
To add a new load file, select Add Load File at the upper left of the tab.
Load File Details Input Fields
Input Field Name | Description | Example(s) |
|---|---|---|
Load Name | This name will be displayed as a card on your sources tab | Production0001_10_10_24 |
Password (opt) | If your compressed file is password protected or encrypted, add your password or decryption |
|
Document Identifier | Enter your unique document identifier column name here | DOCID |
Text Link Path | Enter your text path column name here | text_path |
Encoding - Text | Select encoding type for the text files. (Default is UTF-8) |
|
Encoding - Loadfile | Select encoding type for the DAT file. (Default is UTF-8) |
|
Delimiter - Quote Character | This is the character that encloses content within a DAT file fields. This character will be the lead character when viewing a DAT file with a text editor | System defaults to Concordance standard ASCII 254 character (Greek Thorn) þ |
Delimiter - Field Separator | This is the character that separates content fields enclosed within the quote characters. | System defaults to Concordance standard ASCII 020 character (Box)
|
Delimiter - New Line | This is the line separation character indicating the end of one line and the beginning of another | System defaults to Concordance standard ASCII 174 character ®. NOTE: This character is normally not visible when viewing a DAT file in a text editing program. |
File Type Exclusions - Column Name | Large Language Models do not handle database file types efficiently. This field allows you to exclude file types. Enter your DAT file column name that indicates the file type you want to exclude | FILEEXT |
File Type Exclusions - Extenstion | The extension of file types you would like to exclude from running in Spotlight | csv |
Spotlight AI Management Tab
Spotlight AI is enabled on your matter details page via a “Spotlight AI” toggle. Once enabled, will see a new tab appear on your matter detail page called Spotlight AI Management.
The Spotlight AI management tab will open with a blank table and a Setup button. If you have run Spotlight AI previously, you will see any previous enhancement names in the job table.
Setup Page
Start by selecting Setup
Enter an Enhancement Job Name (required). This name should be descriptive enough so you may understand the enhancement in the future. This name will appear in your enhancement name grid once it has finished running.
Select a Data Source Type (required). In all likelihood you will be selecting your Slack data source.
Enter a case description (required). See case description details below.
Case Description
Your case description essentially defines what Spotlight AI will be looking for when assessing your matter content for relevancy.
It is important to understand that the case description is not keyword based, rather it is using a Large Language Model (LLM) to assess the language contained within your description, and then score your content as relevant or potentially relevant within Illuminate. You may choose to use language directly from a legal complaint, subpoena, or request for production or anywhere where matter information is summarized or put into an executive summary or specifically called out and requested. Alternatively, you may choose to describe your case yourself with any key information or type of language that you may want to focus on for relevancy judgment.
Question Generation
Spotlight AI has multiple ways in which to generate questions that will be run against your data set. You may select Auto-Generate Questions to create an AI-generated list of questions based on your case description. Alternatively, you may select Manual Questions. If you input your own questions, you will bypass the requirement for both the case description as well as the question generation LLM. There is a maximum of 20 questions that you can ask.
Auto-Generate Questions (requires case description)
Spotlight AI has the capability to automatically convert your case description into specific, legal-minded questions. These questions are considered “fire starter” questions meaning they are LLM-generated suggestions that can act as a place to start, but may not be the final question that is run across your data. You have full control over the questions that are run on your content.
Select number of questions you would like Spotlight to generate. Hanzo suggests 10-20 questions.
Select Generate Questions
To add your own question in Auto-Generate mode, select Add Question at the bottom of the question table. This will allow you to input your own question on top of the auto-generated questions.
Illuminate automatically sends your case description to our description LLM whereby it is synthesized and questions are generated for your review. This process will normally occur in less than a minute. Importantly, questions can be edited, removed, or rerun if you choose to modify your case description.
Important note: Your matter data is not processed during the question generation stage!
Editing/deleting questions
You have full control over the questions and may modify, edit or delete the questions at any point in the process. To edit a question, simply select the caret to the left of a question. This will open the question in edit mode. You may reword the question or completely change the question as you see fit.
If you would like to delete a question, simply select the trash-can icon to the right of the question.
Manual Questions
If you would like to bypass the case description/LLM question generation stage, you may use manual mode. Manual mode is best used for internal or HR investigations where case description text may not be available.
Enter Enhancement Job Name and Data Source Type as normal. Navigate to the right side of Spotlight AI settings and select Manual Questions.
Select Add Question
Enter your unique question and select Submit
Repeat the process up to 20 questions
Important note: Avoid asking the same question multiple times. Hanzo recommends unique questions.
Copy to Clipboard
Hanzo is aware that question generation for relevancy may not fall fully within the purview of the classic Illuminate user. For this reason, questions may be generated but need to be communicated to internal or external team members such as in-house or outside counsel or a review team with a trusted legal service provider.
You may select the Copy to clipboard at the top right of the screen thereby copying your questions to the clipboard so that you may paste them into a spreadsheet such as Excel or Google Sheets for collaboration.
Importing CSV to Microsoft Excel
Select the “Copy to Clipboard” from within Spotlight AI
Launch MS Excel
Select or create a Blank Workbook in Excel
Click into Cell A1 to paste data into the top left corner
Select Paste > Use Text Import Wizard from the Clipboard area of your ribbon
From Text Import Wizard -Step 1, select Next
From Text Import Wizard - Step 2, deselect Space and select Comma. Select Next
From Text Import Wizard - Step 3, select Finish
Importing CSV to Google Sheets
Select the “Copy to Clipboard” from within Spotlight AI
Create a new sheet within Google
Select Paste <CTRL V> in Cell A1
Select the clipboard icon at the bottom of your pasted text and select Split text to columns
Save
When you are satisfied with the Spotlight questions, you may select the Save button at the bottom of the screen. You will receive a pop-up message regarding the running of the enhancement job across your selected content.
The Spotlight AI Enhancement job will begin running on your current load file data that has been ingested and the status will show as Running. When the enhancement job is finished, a status of Completed will be shown on the enhancement run.
You may review your Enhancement run or download your DAT overlay file from the Actions button at the right side of your enhance run name.
View Setup allows you to view the original case description and questions for your enhancement run. You may use the copy to clipboard feature to export questions if necessary for auditing purposes.
Download DAT allows you download the Hanzo-generated DAT overlay file.
DAT Overlay File Format
Hanzo produces a DAT overlay file based on your originating, uploaded DAT file. Hanzo will retain your delimiters and produce the following fields as an overlay
Field Name | Description | Example(s) |
|---|---|---|
‘User Defined’ Document Identifier field | This field will replicate the exact docid field identified by the user during the import of the production. | DocID docid Doc_ID Begdoc |
Relevant Rationales | This field has several definitions and is intended to work with Relativities ability to work with multi-layered tags. The first layer is the enhancement job name within Spotlight. The next layer is the question number as shown in Spotlight. The last layer is the actual rationale for why the AI engine flagged the content as relevant. | Enhancement Name\Q10\The document mentions that the employees are paid on a salaried basis. |
Pot Relevant Rationales | This field has several definitions and is intended to work with Relativities ability to work with multi-layered tags. The first layer is the enhancement job name within Spotlight. The next layer is the question number as shown in Spotlight. The last layer is the actual rationale for why the AI engine flagged the content as potentially relevant. | Enhancement Name\Q1\The document mentions specific instances where employees were required to work off-the-clock without being compensated for their time. |
Disclaimer: Spotlight AI is intended as a tool to assist teams with data relevance assessment and should not be solely relied upon for making legal decisions or determinations. Hanzo does not warrant that the tool will be error free, and users should exercise their own professional judgment when deciding on case strategy.