🚀 Loading Coderva...

AI agents for finance and accounting teams

Ledger Insight Analysis helps finance and accounting teams turn paper and PDF paperwork into organized records. You upload bills and receipts, AI extracts the details, and each entry is categorized as Intake, Purchase or Expense. You can then ask conversational questions about revenue, expenses and profit instead of building reports by hand.

Open Ledger Insight

Use cases

  • Upload bills and receipts and let AI extract vendor, date, amounts and line details.
  • Categorize each entry as Intake, Purchase or Expense so records stay consistent.
  • Ask plain-language questions about revenue, expenses and profit.
  • Review extracted data before relying on it, and correct anything the AI misread.
  • Give a small business a lightweight way to keep a working ledger without manual data entry.

Apps and templates that apply

  • Ledger Insight Analysis: the app for uploading documents, reviewing entries and asking questions.
  • Ledger configuration: org admins set the categories the analysis uses.
  • Tools catalog: connect additional tools when an agent needs to reach other systems.

How it works

  1. Open Ledger Insight Analysis and upload a bill or receipt.
  2. The AI extracts the key fields and proposes a category of Intake, Purchase or Expense.
  3. Review the entry, confirm or correct it, and save it to the ledger.
  4. Ask a question such as how expenses compare with revenue for a period.
  5. Use the answers to support decisions, reconciliation and conversations with your accountant.

Who it is for

This domain fits small business owners, bookkeepers and finance staff who handle many receipts and want faster, more consistent records. It supports your accounting work rather than replacing professional judgment, so keep a human review step for anything that feeds tax filings, audits or statutory reporting. A sensible habit is to review extractions at the end of each week, so that errors are caught early and the ledger stays trustworthy for month-end work.

  • Verify before saving: Every extracted field and row is editable.
  • Your own columns: Templates define what is extracted from each document type.
  • Ask your ledger: Answers drawn from your saved bills.

What the finance building blocks give you

Capture paperwork without retyping

Choose a document type, drop in the file and LIA reads it. Scanned images go through vision OCR, PDFs with a text layer are read directly, and Excel sheets are read as tables.

  • JPG, PNG, WEBP, PDF, XLSX and XLS files up to 10MB.
  • Pick the layout first so LIA extracts your exact columns.
  • Or use Create new bill to key in the rows yourself.
  • Live progress while the document is read.
The Upload and Verify tab with a Purchase Register document type selected and a dashed drop area listing the accepted file types

Check every field before it counts

The extracted bill opens in a form. Vendor, invoice number, dates, tax, total and each line-item row can be edited, with a confidence badge and the raw OCR text for reference.

  • Header fields plus a table that uses your template columns.
  • For multi-row registers the total follows the sum of the rows.
  • Nothing is stored until you press Save Bill.
  • A review step you can keep before anything feeds tax filings or audits.
A verification form with vendor, invoice number, bill date, subtotal, tax and total fields, a confidence badge, editable line items with HSN, quantity, rate and GST columns, and Save Bill and Reset buttons

Keep the ledger clean

Each bill gets a category and a status, and saving warns you when the same bill may already be on file. The Bills tab then lets you search and filter everything saved.

  • A warning before saving a bill with the same vendor, invoice number, total and date as one already stored.
  • Categories: Intake (Sales/Fees), Purchase, Expense or Custom.
  • Statuses: draft, confirmed, disputed and paid.
  • Search by vendor, invoice number, notes and OCR text; filter by category, status and date.
The Upload and Verify form with a confirmation dialog warning that a bill from the same vendor for the same amount and date already exists and asking whether to save this one anyway

Ask questions, get answers from your data

Ask LIA about revenue, profit, a vendor, a month or a single entry. It answers from your saved bills and, for register-style documents, from the individual rows.

  • Totals, vendor profiles and period comparisons.
  • Row-level questions, such as the total collected from one class.
  • Short answers with a small table when listing several rows.
  • Says so when the data is not enough to answer.
The Ask LIA tab with a conversation about profit this quarter, a reply with revenue, purchases and expenses, a second question about a class with a table answer, and an input box with a Send button

From a bill to an answer in five steps

The path inside the Ledger Insight Analysis app, from choosing a layout to asking about your numbers.

  1. Pick a document type: Choose a template such as Purchase Register or Expense Report, or let LIA decide with a generic bill.
  2. Upload the file: Add a photo, PDF or Excel sheet. Large files go straight to storage so uploads do not time out.
  3. AI reads the rows: A background job fills your template columns while you watch the progress.
  4. Verify and save: Fix anything that is off, then Save Bill. A duplicate warning appears if the bill is already on file.
  5. Review and ask: Browse the Bills tab, open the Dashboard or ask LIA a question.

Paperwork in, checked records and answers out

Six stages: collect bills and receipts, AI reads the rows, verify and save, organise by category and status, review on the Dashboard, and ask LIA or a Bill Tracker agent, with a strip of review and scope controls

The building blocks for finance and accounting

An app for the team, a ledger configuration that describes your documents, extraction, storage for bills and originals, answers, and agents that can use the same ledger.

Six layers: the Ledger Insight Analysis app, ledger configuration, extraction from images, PDFs and Excel, bill and document storage, question answering, and agents with the Bill Tracker tool
  • Ledger Insight Analysis app: Upload & Verify, Bills, Dashboard and Ask LIA tabs for the whole team.
  • Ledger configuration: Document types, ledger categories and columns for your organisation. Only org admins can change them.
  • Extraction: A background job reads images with Gemini vision, uses the text layer of digital PDFs and reads Excel sheets as tables.
  • Ledger store: Bill records sit in DynamoDB and original files in S3, opened through time-limited signed links.
  • Answers: Ask LIA answers from ledger summaries and line-item rows, and the same data backs the organisation AI assistant.
  • Agents: Attach the Bill Tracker tool in Agent Builder so an agent can save, search and summarise bills in chat.

Screens you will use

Mock-ups with sample data, drawn from the real labels and tabs.

Dashboard

Pick a date range and see revenue, purchases, expenses, profit and bill count, with category totals and trends.

  • Total Revenue, Total Purchase, Total Expense, Total Profit and Bill Count cards.
  • Totals by category and a monthly table.
  • A Download Report button for a printable report.
The Dashboard tab with date range, Refresh summary and Download Report buttons, five metric cards, a totals by category list, a top expenses list and a revenue, purchase and expense trend chart

Ledger configuration

Describe your own documents once. Each document type has a ledger category and a list of columns that drive extraction, the form and item-level answers.

  • Defaults include Sales Invoice, Expense Report and Purchase Register.
  • Colleges and institutes get a Fee Receipt Book instead of a Sales Invoice.
  • Column types are text, number, currency and date.
The Ledger configuration dialog listing Sales Invoice, Expense Report and Purchase Register with their ledger categories and column chips, and Add document type, Reset to defaults and Save configuration buttons

Bill Tracker in an agent

Add the Bill Tracker tool to an agent in Agent Builder and users can upload bills in chat and ask about spending.

  • Operations include create, search, summarize, find_duplicates and mark_paid.
  • Uses PDF Reader on uploaded PDFs.
  • Works on the same ledger as the app.
An agent chat with the Bill Tracker and PDF Reader tools attached, a user uploading a receipt and asking for it to be saved, an agent reply, and chips for create, search, summarize, find_duplicates and mark_paid

How finance teams use it

The same upload, verify and ask loop covers very different documents.

  • A working ledger for a small business: Keep supplier bills, sales invoices and expenses in one searchable place instead of folders and spreadsheets.
  • Weekly review of supplier bills: Use the Purchase Register layout with item, HSN, quantity, rate, GST % and amount, then total purchases by vendor.
  • Expense vouchers: Capture voucher number, paid to, particulars, expense head, mode and approver, and see the top expense items.
  • Fee receipts for schools: Use a Fee Receipt Book layout with receipt number, class and section, and ask about collections by class.
  • Month-end status check: Filter the Bills tab by draft, confirmed, disputed or paid, then Download Report from the Dashboard.
  • Owner questions on the go: Ask how much was spent on fuel last month or what the profit is this quarter, without building a report.

Systems named in the code

The services used for finance and accounting workflows today.

  • Gemini 2.5 Flash (vision OCR, extraction and answers)
  • PDF text layer reading
  • Excel (.xlsx, .xls)
  • JPG, PNG and WEBP images
  • Amazon DynamoDB
  • Amazon S3 (original documents)
  • AWS Lambda and API Gateway
  • Bill Tracker tool in Agent Builder
  • PDF Reader tool
  • Org AI assistant

Review, scope and traceability

Controls that exist in the app today. Keep a human review step for anything that feeds tax filings, audits or statutory reporting.

  • A person confirms every bill: Extraction only fills a form. The record is created when someone presses Save Bill.
  • Duplicate check on save: A bill with the same vendor, invoice number, total and date as one already stored triggers a warning before a second copy is saved.
  • Organisation scoping: Org admins work on the organisation ledger and other users on their own. Only org admins can change the ledger configuration.
  • Original kept and linked: The source file is stored in S3 and opened from the bill through a time-limited signed link.

What is included

  • Vision OCR and text layer: Reads photos, scans and digital PDFs, in batches for long documents.
  • Org-defined columns: Templates control extracted columns, the form and item-level answers.
  • Create new bill: Key in rows by hand when there is no source document.
  • Status tracking: Mark bills draft, confirmed, disputed or paid.
  • Search and filters: Keyword, vendor, category, status and date range.
  • Dashboard and report: Revenue, purchase, expense, profit and a downloadable report.
  • Row-level answers: Ask about one person, receipt or voucher, not only totals.
  • Bill Tracker for agents: Save, search and summarise bills from an agent chat.

Frequently asked questions

What kinds of documents can I upload?

You can upload bills and receipts, and the AI extracts details such as vendor, date and amounts. Review each extraction, because image quality and layout affect accuracy and some fields may need correction.

How are entries categorized?

Each entry is categorized as Intake, Purchase or Expense. The AI proposes the category from the document, and you can confirm or change it so that your ledger reflects how you actually classify transactions.

Can it replace my accountant?

No. It speeds up data capture and helps you explore your numbers, but accountants remain essential for compliance, tax advice and final reporting. Treat the output as a working aid that you verify.

Related

Open Ledger Insight