Skip to content
Kuberan AI Help

AI Agent

Open AI Agent

Overview

The AI Agent is a conversational assistant embedded within each client’s accounting file. It reads the client’s transactions, invoices, bills, balances, and reports to answer questions in plain English — without requiring you to run a report or filter a list manually. Beyond answering questions, the agent can create transactions on your behalf (with your confirmation), surface anomalies and trends in the books, and manage the queue of documents waiting for AI data extraction. The agent’s context is scoped entirely to the current client file; it cannot access data from other clients in your workspace.

AI Agent screen showing chat interface on the left, document processing queue on the right, and AI suggestions panel below
The AI Agent screen with the chat interface, document processing queue, and proactive suggestions panel.
  1. (1) Chat input — type a question or instruction in natural language
  2. (2) Chat history — conversation thread with data references and links
  3. (3) Document Processing Queue — uploaded documents awaiting AI extraction
  4. (4) AI Suggestions panel — proactive flags and recommendations
  5. (5) Confirm Transaction button — approve an agent-proposed transaction before posting

Screen Layout

The AI Agent screen has three zones:

Chat panel (left/centre): The main conversational interface. Your messages appear on the right; agent responses appear on the left. Responses often include embedded data tables, chart account names (as clickable links), and a Show Source toggle that reveals the underlying query or data used to generate the answer.

Document Processing Queue (right, top): A live list of documents uploaded to this client file that are currently being processed by the AI extractor or are awaiting your review. Each entry shows the file name, upload time, and AI status.

AI Suggestions panel (right, bottom): Proactive recommendations generated by the agent after analysing the books — for example, uncategorised transactions that match a vendor pattern, an invoice approaching 60 days overdue, or an unusual expense spike.

Chat Interface

What You Can Ask

The AI Agent can answer questions about any data in the current client’s books. Example queries:

  • “What was net income for Q3 2024?”
  • “Show me all expenses over $5,000 in the last six months.”
  • “Which invoices are more than 60 days overdue?”
  • “What is the current GST/HST balance owing?”
  • “How much did we spend at [vendor name] last year?”
  • “Is there anything uncategorised from last month?”
  • “What is the largest expense category year to date?”
  • “Compare revenue this quarter versus last quarter.”

The agent draws on the same data as the Reports and Bank Transactions screens. Answers include the figures requested and are accompanied by references to the source accounts or transactions.

Creating Transactions via Chat

The agent can draft transactions based on natural language instructions. This is useful for quick entries when you are reviewing the books conversationally.

Examples:

  • “Record a $350 expense to Office Supplies for a purchase on April 10 paid from the RBC chequing account.”
  • “Create a $1,200 journal entry to record prepaid insurance for May.”

The agent will present a draft transaction card showing the proposed date, accounts, amounts, and description. You must click Confirm before the transaction is posted to the ledger. If the proposed entry looks wrong, click Edit to modify it or Discard to cancel.

UI Elements

Chat

Chat Input

Free-text field at the bottom of the chat panel. Press Enter or click the Send button to submit. Shift+Enter inserts a new line for multi-sentence queries.

Show Source

Available on each agent response. Expands to show the accounts, date ranges, or transaction filters the agent used to calculate the answer. Useful for verifying accuracy before acting on a figure.

Confirm Transaction

Appears on agent-proposed transaction cards. Clicking posts the transaction to the ledger. The button is labelled with the transaction type (e.g., Post Journal Entry, Post Expense).

Edit

Appears alongside Confirm on transaction cards. Opens the relevant transaction editor (journal entry, bank transaction, etc.) pre-populated with the agent’s proposed values so you can adjust before posting.

Document Processing Queue

Queue Item

Each row in the queue shows the document file name, the time elapsed since upload, and the current AI status (Queued, Processing, Review Required). Click Review to open the AI Extraction Review panel for that document.

Review All

Opens a batch review flow that steps through all documents with a Review Required status one by one, allowing you to confirm or correct the extracted data and post transactions rapidly.

AI Suggestions Panel

Suggestion Card

Each card describes a finding and includes a one-click action button. Examples: “3 uncategorised transactions from [vendor name] — click to apply suggested category”, “Invoice INV-0047 is 62 days overdue — send reminder”, “Unusual expense spike in Travel this month vs. prior 3-month average”.

Dismiss

Dismisses a suggestion card. Dismissed suggestions do not reappear unless the underlying condition changes.

Actions

Ask a Question

  1. Click in the Chat Input field at the bottom of the chat panel.
  2. Type your question in plain English (Canadian English is fully supported — “cheque”, “GST/HST”, “T4A” are all understood).
  3. Press Enter or click the Send button.
  4. The agent responds within a few seconds. Review the answer and click Show Source to verify the underlying data if needed.
  5. You can follow up with clarifying questions or ask for more detail. The agent maintains context for the current conversation session.

Create a Transaction via Chat

  1. Type an instruction describing the transaction you want to create (e.g., “Record a rent payment of $2,500 to ABC Properties on March 1 from the TD chequing account”).
  2. The agent proposes a transaction card with the populated fields.
  3. Review the proposed account, amount, date, and description.
  4. Click Confirm to post, Edit to adjust, or Discard to cancel.

Process Documents from the Queue

  1. In the Document Processing Queue, click Review All or click Review on an individual document.
  2. The AI Extraction Review panel opens. Verify the extracted fields and correct any errors.
  3. Click Post Transaction to create the ledger entry, or Dismiss to skip.
  • Documents — Upload documents that feed the AI processing queue
  • Bank Transactions — Categorise transactions surfaced by AI suggestions
  • Journal Entries — Agent-created journal entries appear here
  • Reports — Run full reports for data that goes beyond a single chat question