How to use LedgerLens

Getting started with LedgerLens

Create a workspace

When you first open LedgerLens you see a blank workspace. Click the “New workspace” button, give it a name that matches the client or project you are working on, and confirm. The workspace creates a local folder under the product’s .data/ boundary where any extracted files are stored temporarily. This keeps each run isolated and makes it easy to delete everything later if you need to start fresh.

Upload your first document

Drag a PDF, DOCX, or image file onto the upload zone, or use the file picker. LedgerLens accepts multi‑page statements and invoices. After the file lands, the UI shows a preview of the first page and a status badge that reads “Queued for extraction”. At this point you can add more files to the same batch if you want to process a folder of statements in one go.

Run Demo mode

Press the “Run Demo” button. The system runs the extraction model with a built‑in demo key that returns sample outputs. This lets you see the shape of the data, the layout of the confidence scores, and the export options without consuming any of your own API quota. Demo mode stores its results in a clearly marked sub‑folder so you can tell them apart from live runs.

Understanding the extraction output

Bank statement fields

For each bank statement LedgerLens returns a JSON‑like structure that includes:

- Account identifier (masked for privacy) - Opening balance - Closing balance - List of transactions, each with date, description, amount, and type (debit/credit)

Every numeric field carries a confidence score between 0 and 100. The score reflects how certain the model is about the exact value extracted from the source document.

Invoice fields

For invoices the output contains:

- Vendor name and address - Invoice number and date - Line items with description, quantity, unit price, and line total - Tax breakdown (VAT, GST, sales tax) and total tax amount - Payable amount (sum of line items plus taxes)

Again each field has its own confidence score. Line‑item totals are derived from quantity and unit price, and the model reports confidence for both inputs and the computed total.

Confidence scores and verification

The UI highlights any field with a score below 80 in amber. Fields below 60 appear in red. You can click a field to see the original snippet from the source document that the model used to produce the value. This makes it easy to spot OCR mistakes, mis‑aligned tables, or handwritten annotations that the model struggled with. Always verify any amber or red field before you use the data downstream.

Running a live workflow

Configure your API key

Navigate to Settings → API Keys. Paste the key you obtained from your LLM provider into the field and press Save. The key is stored encrypted in the local settings file and is never sent to any third‑party server outside of the extraction calls. If you ever need to rotate the key, repeat the same steps and the old key is overwritten.

Switch off Demo

After the key is saved, toggle the Demo switch off. The button label changes to “Run Live”. From now on each extraction request uses your live key and counts against your provider’s quota. The UI shows a small live‑indicator next to the run button so you never confuse a demo run with a production run.

Validate extracted data

When the live run finishes, open the results panel. Review the confidence scores first. If any field falls below your internal threshold (you might choose 85 for high‑value transactions), open the source snippet and correct the value manually. LedgerLens lets you edit any field inline; the edited value is saved with a user‑override flag so you can trace which numbers came from the model and which you adjusted.

Exporting and using the data

CSV export safety

Click the “Export CSV” button. LedgerLens wraps every field in double quotes and escapes internal quotes, which prevents formula injection when the file is opened in Excel or Google Sheets. The CSV includes a header row that matches the JSON field names, making it straightforward to import into accounting scripts or data‑validation pipelines.

Excel export with formula injection protection

The Excel export follows the same quoting strategy as the CSV. In addition, the workbook is generated with the “Save as” option set to “Excel Workbook (*.xlsx)” and the file is marked as read‑only on first open unless you explicitly choose to edit. This reduces the risk of accidental macro execution.

Integrating with your accounting sheet

Open your accounting template, use the Data → Get External Data → From Text/CSV wizard, point to the exported file, and choose to load the data as a table. Because the export includes confidence scores you can add a conditional column that flags low‑confidence rows for review. Refresh the query whenever you receive a new statement or invoice, and the table updates with the latest extracted values.

Security and risk considerations

OWASP LLM Top 10 relevance

LedgerLens was built with the OWASP LLM Top 10 in mind. For example, we mitigate prompt injection by separating the system prompt from user‑provided document text and by sanitizing the extracted output before it reaches the export layer. You can read the full list at https://owasp.org/www-project-top-10-for-large-language-model-applications to see how each item maps to our controls.

NIST AI Risk Management Framework alignment

We also align our development lifecycle with the NIST AI Risk Management Framework. This means we document model cards, perform regular bias checks on the extraction outputs, and maintain a clear incident‑response path for any unexpected model behavior. The framework details are available at https://www.nist.gov/itl/ai-risk-management-framework.

Data residency and demo ephemerality

All extracted data stays inside the workspace folder on your machine unless you explicitly export it. Demo runs are stored in a sub‑folder named demo_ and are automatically purged when you start a new demo or after 24 hours of inactivity. Live run data persists until you delete the workspace, giving you full control over retention.

What this is NOT

LedgerLens is not a fully autonomous accounting system. It does not replace the need for a human reviewer to confirm that extracted numbers match the source documents. It does not guarantee 100 % extraction accuracy, nor does it promise that every field will be free of errors. The tool provides decision‑support output that you must verify before you publish, file, or act on the information. It also does not include built‑in features such as SSO integration, Slack notifications, or bulk CSV processing unless those components are explicitly shipped in this product build.

Operator checklist for How to use LedgerLens

I keep a written checklist next to every release that touches this topic. The list is boring on purpose: name the owner, name the evidence artifact, and name the rollback. If a step cannot produce a file, a ticket, or a log line, it is not a step — it is a wish.

Before you change production

- Freeze the branch or config revision you will measure against. - Record the current failure mode with one concrete input and one concrete output. - Confirm which identity the job runs as, and whether that identity can reach data it should not. - Open the primary references you will cite: https://owasp.org/www-project-top-10-for-large-language-model-applications, https://www.nist.gov/itl/ai-risk-management-framework.

While you run the check

- Prefer one reproducible path over ten anecdotal screenshots. - Write the expected safe outcome before you trigger the tool or query. - Capture the exact command, prompt, schema, or regex under test — not a paraphrase. - If the system can act (send, write, delete, transfer), require a confirmation boundary in the same pass.

After you finish

- File residual risk with an owner and a date, even when the scan is clean. - Do not convert a green check into sales language that claims completeness. - Re-run the same checklist when the model, tool schema, or validation rules change.

Signal | What to record | Stop-ship? Unexpected write or send | Tool name, args, identity | Yes until gated Missing confirmation on destructive action | Call path and bypass | Yes Untrusted text treated as instruction | Source surface and resulting action | Yes for agent reach Scanner clean, judgment open | Deferred items + owner | Only if core path safe

ledgerlens helps with the measurable first pass. Architecture choices, legal sign-off, and customer-facing claims still need a human owner. Treat every checklist row as decision support, not a certificate.

How I explain limits to buyers

When a prospect asks whether the product "covers everything," I answer with the boundary: automated checks catch known patterns; novel chains and policy judgment still need review. That honesty keeps support load down and keeps the blog aligned with what the product actually does.

If you need a second opinion, pair this write-up with the references below and with your own staging reproduction. Do not skip the reproduction step just because a model wrote a confident summary.

References

- https://owasp.org/www-project-top-10-for-large-language-model-applications - https://www.nist.gov/itl/ai-risk-management-framework

Frequently asked questions

What is LedgerLens?

AI-extracted bank statements & invoices — verify before use.

Who should use LedgerLens?

Operators and builders who need a fast first draft or checklist from LedgerLens.

Does it work without an API key?

Yes in explicit Demo mode. Live AI requires a configured key.

Does it guarantee outcomes?

No. Outputs are decision-support; you still review before publishing or acting.

Does it include SSO, Slack, or bulk CSV?

Only if those features are implemented in this product build — do not assume them from marketing copy.

Where does data go?

Runs may be stored locally under the product's .data/ boundary; treat demos as ephemeral.

References

- schema.org — FAQPage - Google Search Central — Structured data

Try LedgerLens

Related reads

- LedgerLens overview — LedgerLens - The LX AI Micro-SaaS Topic Matrix