Agentic Table

12 min read

Overview

The Agentic Table is flexible, spreadsheet-like interface powered by Unique AI to provide a structured interface for expediting frequent and repetitive tasks.

Who is it for

Agentic Table has a wide range of use cases that can support the entire organization at financial services firms. Customers have found Agentic Table use cases across investment teams, front and back office, leadership, and more.

If you’re unable to access certain features or sections of this article, it’s possible that your firm doesn’t have access or hasn’t upgraded to the latest version. Please reach out to your internal support team for further assistance.

Benefits

Batch AI Insights

• Automate the population of your tables with Unique AI
• Store pre-approved answers in the Answer Library to ensure consistent content and styling

Collaborative Platform

• Share tables with the entire team and perform simultaneous edits
• Row verification and locking features ensure data integrity and allow teams to track progress

Flexible Data Extraction

• Import and export tables from various file formats (Excel, CSV, PDF, Word).
• Leverage Unique's Knowledge Base to find answers from many data types (Excel, CSV, PDF, Word, Sharepoint, PPTX, TXT, HTML, MD)

Use Cases

Agentic Table is highly extensible and can be used for a variety of general as well as finance specific use cases.

Structured data extraction - Use Agentic Table to extract structured data from all your Knowledge Base data

Bulk Query submission - Prepare extended sets of questions which can be answered asynchronously

RFP/DDQ - Answer incoming RFP/DDQs and prepare new standard answers for prospective investors and clients

Due Diligence - Perform research and answer investment committee questions using Unique AI to gather information and deliver tailored responses

KYC - Verify the identity of your clients and assess potential risks using standard questionnaires and existing documents

Compliance/DORA - Complete financial and IT compliance forms from internal and external teams

Step-by-Step Guide

1. Landing page

The landing page provides the main entry point to creating a new sheet. It displays all sheets in a searchable table. From this interface, users can create, open, rename or delete sheets.

2. Create Sheet

Users are guided through three simple steps to create a new sheet with extracted questions and AI generated answers.

General Information

Give your sheet a name and provide a due date (optional).

Import Questions

Upload a question file if you'd like to import a set of questions from a questionnaire (e.g., DDQ), meeting minutes, or a list of KPIs.

Users will be prompted to upload or select from Knowledge Base a question file

Users can add question files from the Knowledge Base document selection panel or their local file system (seen above)

Add Sources

Specify a set of sources for knowledge to answer the questions

Users will be prompted to select source to search for answer from their Knowledge Base

Users can add question files from the Knowledge Base document selection panel (seen above) or their local file system

Only supported file types are compatible: Please note that only specific file types can be utilized for question extraction and as part of the knowledge base. For more details, visit: Supported File Types

It is also possible to skip Import Questions and/or Add Sources steps and start with only question or a blank sheet.

Review file selection and create your table

Click “Next” to create your sheet

Question files containing “_export” in the title will be imported as a completed sheet without AI answering. Columns will be imported to match the existing space configuration - columns missing will be left blank and extra columns will be dropped.

3. Edit your Sheet

Inside the table UI, users are enabled to perform a number of actions:

1. Edit cells content

2. Double click or use the Enter key to open cells in editing mode

3. Click outside of the selected cell or use the Enter key to submit cell changes

4. All cells support full markdown

5. Trigger agentic workflows and generate content for other cells. (eg. Editing a question will call the agent and generate a new answer, references, and hallucination check.)

6. Add further knowledge sources, question files or individual questions.

7. View a list of all files that are associated with the sheet.

8. Export sheets as DOCX or CSV reports Agentic Table

9. Review references by clicking on provided links.

10. Review any cell’s change history to monitor who edited a cell and when (both AI or human).

11. Delete rows

4. Share your Sheets

Users who create a sheet are deemed owners of that sheet. As the owner of a sheet, you can choose to share that sheet with other members of the space. Users be granted one of two roles “Can manage” or “Can edit”. The permissions associated with these roles are detailed in the table below.

Can answer is a restricted collaborator role for contributors who should only work on the rows assigned to them. A Can answer user can see, manually edit, and comment on only the rows they are assigned to (plus the header row) — all other rows are hidden from them. They cannot trigger AI generation or re-run cells, add or delete rows, add questions or sources, change sheet-level settings (due date, table instructions, metadata, or completion status), or rename, share, or delete the table.

⚠️ This feature is still experimental and is gated behind the feature flag FEATURE_FLAG_AGENTIC_TABLE_CAN_ANSWER_ROLE_UN_21308, which is disabled by default.

Private Sheet

Shared Sheet

Sharing a sheet is a three click process:

1. Click the Share button to open the Sharing modal

2. Type in the search bar for users or groups to grant permissions

3. Select the desired permission level from the dropdown

4. Click “+ Add” to submit your sharing updates

If you do not see the search bar to add other users, you likely do not have permissions to share the sheet with others. Contact the sheet owner or manager to request additional permissions

5. Due Date

Due dates allow users to track sheet progress and set deadlines for sheet completion. Sheets past their due date will be marked Overdue but remain all editing capabilities. A sheet may be given a due date at time of creation or any time after when viewing the sheet. Due dates are optional.

Adding due dates

Users will find the due date option in the ‘General’ section of the New Sheet creation flow. Users can click Add Due Date to assign a due date to their new sheet.

Users can find the due date displayed in the Agentic Table toolbar. Clicking on the due date badge will display a calendar where users can edit the due date

Viewing due dates

Due date status can be seen in the “My Sheets” section of a space and the Agentic Table toolbar. Overdue sheets are marked with an Overdue badge.

My Sheets Table

Due dates as seen in “My Sheets”. Note: overdue sheets are marked with an Overdue badge in the due date column.

Agentic Table Toolbar

A on-time completed sheet as seen in the Agentic Table toolbar

An overdue sheet as seen in the Agentic Table toolbar.

6. Review Status

Users can create Review Status columns that enhance workflow tracking and implement restriction on agent and manual cell editing. These columns come with a default configuration, but also offer the option to customize the review status process to the user’s workflow.

Review Status Disabled

Users can configure their space to have no review statuses. To do this, add this configuration to Advanced Settings.

"magicTableConfig": {
  "hideSheetStatus": true
}

Default Review Statuses

The default review status configuration allows users to mark rows as Needs Review, Ready for Verification or Verified. This column and set of statuses also allow users to control or lock the ability to edit the table. The expected behavior is as follows:

• Needs Review - Rows marked Needs Review are fully unlocked and can be edited by the AI agent and manually by users

• Ready for Verification - Rows marked Ready for Verification can be edited manually by users but are locked from editing from the AI Agent

• Verified - Rows marked Verified are locked from any editing

Review Status column displaying an opened Review Status dropdown and all three values when selected

Custom Review Statuses

In addition to the default statuses, users can define a custom review status list to tailor review workflows to their specific use case. Custom statuses allow configuration of number of steps, names, colors, and editing permissions. For more details on configuring and using custom review statuses, see Configure Custom Review Status.

7. Assign Users

The assignee column allows users to direct collaborators to questions that require their attention. Users can assign a question to anyone who has access to the sheet. If a sheet is private, the owner can only assign questions to themselves; whereas if a sheet is shared, they may assign it to anyone in the space. A row may be assigned to multiple users.

Assignee column displaying an unassigned row, unassigned row with opened assignee selector, and row assigned to Alex Albracht

8. Export your Sheet

Users can export the data contained in the Agentic Table as a PDF report or complete excel.

List of Exports supported:

• Full Report - (DOCX): This export will generate a word document using a customizable template. the standard export contains the questions, answers, references, and hallucination checks from an agentic table.
• Questions: This export will generate a word document using a customizable template. This export contains only a numbered list of the questions from an agentic table.
• Full Report (XLSX): This export will create an excel spreadsheet with the agentic table loaded into a single sheet. All columns will be included in the export. References within the answer column(s) are simplified to use only the reference number.
• Full Report (XLSX) - Unformatted: This export will create an excel spreadsheet with the agentic table loaded into a single sheet. All columns will be included in the export. References within the answer column(s) include the full message_id and reference_id. This allows users to reimport as a new sheet and preserve the reference links.

Selecting the “Export” button opens a dropdown where users can select their desired export format

Microsoft Word documents will prompt a user to download when the file is available. Excel files will automatically download into the browser.

9. Feedback

The Agentic Table supports feedback on LLM-generated cell content, similar to the feedback functionality available in the chat interface. This enables consistent review and quality control of AI-generated outputs at the cell level.

To submit feedback on a cell:

1. The user right-clicks on any cell populated by an agent LLM call.

2. The user selects Feedback from the dropdown menu.

3. The user enters the feedback and submits it.

Submitted feedback is included in the feedback export, consistent with the handling of chat message feedback.

Unlike the chat interface, cells can be re-generated, so users may submit multiple feedback entries for the same cell. Each entry is linked to the cell’s specific generated version.

ServiceNow Integration

Agentic Table cell feedback can be connected to ServiceNow.

More details on configuration and setup are available on the dedicated ServiceNow integration page.

10. Table Metadata & Instructions

Agentic Tables support Sheet-Level Metadata and Table Instructions to guide the AI agent’s processing of table content.

• Sheet-level metadata allows tagging and organizing data to enhance filtering, search, and retrieval.
• Table Instructions provide optional descriptive instructions that inform the AI agent, improving the relevance, clarity, and accuracy of LLM-generated responses.

These features enable structured data management and guide LLM output for more accurate and relevant responses.

Sheet-Level Metadata

Sheet-level metadata is defined during sheet creation and applied to all rows within the sheet. It consists of key:value pairs, separated by semicolons.

Key Features:

• Metadata input is optional and applied to all rows atomically.
• Entries are semicolon-separated key:value pairs (e.g., department:sales; region:north; priority:high).
• The metadata section on the sheet creation page is collapsible by default to maintain a clean interface.
• Metadata can be edited after sheet creation, but changes only apply to rows generated afterward; existing rows retain the metadata that was originally applied.

Table Instructions

Table context is an optional free-text description associated with the table that provides instructions or contextual information to the AI agent. It can be dynamically inserted into prompts using Jinja templates (e.g. {{ table_context }}).

Key Features:

• Optional text input during table creation or later updates.
• Excluded from system prompts if left empty, preventing unnecessary clutter.
• Table Instructions can be edited after sheet creation, but changes only apply to rows generated afterward; existing rows retain the instructions that were originally applied.

Table instructions and metadata can be added during sheet creation.

Table instructions and metadata can be displayed and edited on the sheet page.

11. Comments

Agentic Table supports cell-level commenting so collaborators can leave notes, ask questions, and tag teammates directly on the data — without modifying the underlying cell content.

Leaving a comment

1. Right-click a cell and select Comment (or open the cell and use the comment action).

2. Enter your comment. Use @ to tag (mention) another user with access to the sheet.

3. Submit to post the comment.

Cells with active comments display a small visual indicator so they are easy to spot in the table.

Comment modal

Clicking a commented cell — or a comment in the side panel — opens the comment modal with the related cell selected and highlighted. From the modal users can:

• Read the comment and any replies in the thread
• Reply to the thread
• Edit or delete their own comments
• Resolve the comment when it has been addressed
• Copy a direct link to the comment for sharing

Comments side panel

A Comments button in the toolbar opens the side panel, which provides an aggregated view of all comment activity on the sheet. The panel is organized into three tabs:

• For You (default) — comments where you have been mentioned
• All comments — all active (unresolved) comments on the sheet
• Resolved — resolved comments, shown in a grayed-out style

Each entry shows the author, comment text, timestamp, the cell it refers to, and any replies. Clicking an entry navigates to the cell in the table and opens the comment modal in context.

Permissions

All users with access to the sheet (Owner, Can manage, Can edit) can leave, reply to, and resolve comments. Editing and deleting are restricted to the comment's author. See the permission matrix in Share your Sheets for details.

Analytics

A list of currently tracked metrics can be found here.

Data Retention

Agentic Tables support an optional data retention policy that automatically deletes old sheets and any manually uploaded content after a certain number of days. The cleanup runs once a day at midnight.

Data retention can be managed either globally (for all spaces) or per space:

• A global setting defines how long (in days) data is kept overall.
• A space-level setting can define a custom retention period — but only if the global option is enabled or set to a specific number of days.
• Files uploaded to the Knowledge Base are NOT impacted

Please note that data retention can’t be used for spaces with an Answer Library, as these features are mutually exclusive.

For a full configuration guide, see the technical documentation.

Tips & Tricks

Keyboard navigation: Users can navigate the table using their keyboard’s arrow keys and use the enter key to enter and exit cells for easy editing.

Locking limits edits: When the agent interacts with the table, the table will be locked from user editing. When the agent is editing the sheet, the sheet will present an “Updating…” state and users will be unable to click into any cells.

Limitations

Sharing across spaces: Sheets cannot be shared outside of a space. Future enhancements will transition sharing to a user level permission rather than space level.

Answer Library AI restrictions: Answer Library sheets are not enabled with AI. This has been implemented by design to protect user Validated answers.

Architecture