Tasks AI Agent

How to work with DigiFi's Tasks AI agent, a built-in assistant that automates previously-manual tasks.


Overview of the DigiFi Tasks AI Agent

The DigiFi platform features a built-in AI assistant that helps your team streamline the review and completion of application tasks using generative AI.

Once enabled, the Tasks AI Agent becomes available within application tasks, allowing you to review documents, update data, and complete tasks more efficiently. It can perform actions such as reading and verifying documents, extracting and updating variables, assigning labels, performing calculations, and adding comments.


Tasks AI Agent Availability and Permissions

The Tasks AI Agent is available within application tasks and can also be triggered automatically through Automation Builder.

  • The Agent operates with the same permissions as the user working with it. For example, if a user does not have access to certain data, the Agent cannot access it either.
  • Multiple users can collaborate with the Agent on the same tasks.

Tasks AI Agent File Upload Requirements

Supported File Formats

The Tasks AI Agent supports the following file formats: .pdf, .json, .csv, .txt, .xml, .svg, .heic, .png, .jpg, .jpeg, .gif, .webp

File Size Limits

The Tasks AI Agent has file size limits, most of which are requirements of the underlying models. DigiFi allows files up to 25MB, and then performs compression prior to sending files to the AI Agent.

If the file is not smaller than 4.5MB, or cannot be compressed below 4.5MB, then the agent may be unable to process the file. In those cases, automated processing will be paused until a smaller file is provided.


How the Tasks AI Agent Works

When working with a task, the AI Agent goes through three phases:

  • Planning – Creates a plan based on your internal instructions.
  • Processing – Reviews documents, analyzes data, and performs updates.
  • Completion – Summarizes the work done and records the results.

The Tasks Agent begins by reading your internal instructions and generating a structured plan. These steps can include actions such as:

  • Reading task data (title, internal instructions, variables in the data tab, documents, labels, status)
  • Updating task data (labels, status, variables)
  • Reading and analyzing documents (verifying authenticity, extracting data, matching with existing data)
  • Adding task comments (and reading existing ones)
  • Reading some application data (status name, application labels, application variables and IDs)
  • Performing calculations based on available information and updating variables

During Processing, the Agent executes the plan step by step and logs each action. For example, when analyzing a document, it displays the document name and ID and provides verification results if requested.

Every update completed by the agent is marked with a special icon in the application history, making it easy to track.

  • If the history record shows API, the action was executed automatically (via workflow automation).
  • If it shows a user name, the action was performed manually in guided mode.

Tasks AI Agent Execution Modes

The Tasks AI Agent supports two execution modes:

Guided Mode

In guided mode, certain actions require user confirmation before being applied. Users can confirm or decline suggested updates and provide feedback if adjustments are needed.

The Agent may pause and request guidance if instructions are unclear, critical data is missing, or business judgment is required.

Automatic Mode (Default)

Users can enable the “Run Automatically” toggle to allow the AI Agent to execute all internal instructions within a task without user confirmation.

In automatic mode:

  • The Agent executes defined tasks without step-by-step confirmations.
  • Confirmation-related blocks are hidden during execution.
  • All actions are logged.
  • A complete execution history is available after processing.
📘

To automate Tasks AI Agent in your workflow tasks you can use the “Run AI Agent” action in Automation Builder.

Place Run AI Agent actions at the end of your workflow.

The AI Agent runs longer than other automated actions and may take 1-2+ minutes to complete.
If subsequent workflow actions depend on its output, add a Time Delay to ensure processing is complete before the workflow continues.


Best Practices for Writing AI Agent Instructions

To get the most out of the AI Agent, follow these guidelines when writing instructions:

  • Clear Instructions: Give direct instructions with clear rules. The AI Agent needs to understand exactly what you want it to do, with precise guidelines.
  • Specify Formats: If the agent needs to save a value, clearly state the format (e.g., MM/DD/YY) to ensure compatibility with the variable.
  • Use Full Variable Names: When extracting data, specify the variables where the data should be saved.
  • Separate Steps: Use multiple, clearly separated steps (e.g. 1, 2, 3) to provide the agent ordered, multi-step instructions.
  • Use Conditional Instructions: Create “if-then” rules within your text where needed. For example: if the document seems suspicious, assign a label and update a variable.
  • Detail Calculations: When instructing the agent to perform calculations, specify which values to use and the formula. For example: “If pay frequency and gross pay are available, use: gross pay × frequency multiplier (52, 26, 24, or 12).”
👍

❌ Bad: “Check the ID.” ✅ Good: “Complete the following steps to help verify the applicant's identity:

  1. Validate the identity document to determine if it appears to be either a passport or driver's license, authentic, not manipulated and not expired. If the document is the wrong type, or appears to be fake, manipulated or expired, assign the task label "Invalid Document", provide an explanation in the Task Comments and update the Task's status to "Failed".

  2. Compare the first name, last name, address and date of birth in the identity document to the values in the variables Borrower First Name, Borrower Last Name, Borrower Home Address and Borrower Date of Birth. If they are not the same or very similar, assign the task label "Data Mismatch", provide an explanation in the Task Comments and update the Task's status to "Failed".

  3. If all steps are completed without issues, change this Task's status to "Done""


Important: Handling Missing Documents or Data

If the AI Agent cannot find required inputs (such as document or label referenced in the instructions), it will pause execution and request additional input – in both guided and automated runs.

When triggered via Automation Builder, the Agent may stop processing and wait for the missing information before continuing.

Please note:

  • If a document referenced in the instructions is missing, the Agent will request it.
  • If a label does not exist, it cannot be assigned.
  • If a variable is not linked to the task, it cannot be updated.

Behavior depends on how the instructions are written. If the instructions allow skipping missing data, the Agent will proceed accordingly.

Clearly define how to handle missing documents, labels, or variables in your task instructions to avoid interruptions.


Tasks AI Agent Billable Actions

The AI Agent is billed based on the actions it performs while processing tasks. Starting the Agent does not itself count as a billable action. Only eligible operations performed by the Agent are charged.

One billable Tasks AI Agent action equals 10 AI Credits.

For general information about AI Credits, usage limits, and billing notifications, see AI Credits.

Billable Actions:

  • Document analysis: Each page of a document that the Agent is instructed to read or analyze counts as one billable action. For example, analyzing a 10-page document counts as 10 billable actions.

  • Variable updates: Updating task variables counts as a billable action. The number of billable actions depends on how the updates are grouped. Reading variables or calculating values without updating them is not billable.

  • Comments: Adding a comment to a task counts as one billable action. Reading existing task comments is not billable.

  • Labels: Adding or removing a task label counts as one billable action per.

  • Status changes: Changing a task's status counts as one billable action.

The way instructions are structured can affect how variable updates are grouped. For example, asking the Agent to update several variables in a single grouped operation may result in one billable action. Instructing the Agent to update each variable separately may result in a separate billable action for each update.