How long does a first audit take?

Uploading and extracting a year of bank statements usually takes a few minutes per document. The audit itself runs over the whole population in seconds once the ledger is normalised.

Do I need my accounting software connected?

No. NanoFora works from the documents themselves — statements, exports, invoices. Connecting a system is optional and only saves you the upload step.

Where do the credits go?

Credits are consumed by AI work: document extraction, audit runs, assistant answers and report generation. The cost is shown before you confirm, and the balance and history are on the Billing page.

Can I trust the forecast numbers?

The projection is deterministic maths applied to your own history and your stated assumptions — not a model guess. Every scenario is backtested against your real months so you can see the error range before relying on it.

Is this a statutory audit?

No. NanoFora produces analysis and exception testing for management use. Statutory accounts, tax filings and formal audit opinions must come from a qualified professional.

Help Centre — How NanoFora Works, Step by Step

Help centre

How NanoFora works, from a folder of statements to a forecast you can defend

Follow the eight stages below in order. Each one explains what you do, what the platform does with it, and what you end up holding at the end.

Start a workspace Read the FAQ

The flow at a glance

01
Create your workspace
02
Upload your financial documents
03
Extraction and review
04
Normalisation into double-entry accounting
05
Run the AI audit
06
Work the findings
07
Reports and future assumptions
08
Export, share and monitor

Step 01

Create your workspace

Every company is an isolated workspace with its own data, members, currency and permissions.

Sign up and complete onboarding: company name, country, base currency and financial year.
Invite your team by email and assign a role — owner, admin, CFO, accountant, analyst, manager, employee or viewer.
Set the workspace page size, logo and notification preferences in Settings.

You end up with: An empty, access-controlled ledger ready to receive documents.

Step 02

Upload your financial documents

NanoFora reads the files you already have — no template, no manual re-keying, no accounting software required.

Drag in bank statements (PDF), invoices, receipts, or exports in CSV, XLSX, OFX and QFX.
Select multiple files at once; each is queued with a persistent processing status you can leave and come back to.
The credit cost of the run is shown before you confirm, and submission is blocked if your balance is short.

You end up with: A document library with page-level source files retained for evidence.

Step 03

Extraction and review

Each document is parsed into structured transactions with a confidence score on every field.

Open the extraction preview to see dates, descriptions, amounts, currency and counterparty side by side with the source.
Re-map any column the parser read differently from your bank's layout.
Anything below the confidence threshold is flagged for review instead of being accepted silently.

You end up with: Confirmed transactions, each linked back to the page it came from.

Step 04

Normalisation into double-entry accounting

Raw lines become a proper ledger: categorised, signed correctly and mapped to a chart of accounts.

Transactions are classified into income, expense, asset, liability and equity accounts.
Foreign-currency lines are converted to your base currency using dated rates, with the original amount preserved.
Opening balances and inter-account transfers are handled so cash does not double-count.

You end up with: A balanced ledger that a P&L and cash-flow statement can be built from.

Step 05

Run the AI audit

Eleven deterministic checks test the whole population, not a sample, against your materiality settings.

Set materiality (amount or % of turnover), the round-number multiple, outlier sensitivity and the duplicate window.
Enable or disable individual checks, then run the audit for a chosen period.
Schedule it to repeat monthly or quarterly so exceptions surface without anyone remembering to look.

You end up with: A findings list with severity, confidence, exposure and an opinion for the period.

Step 06

Work the findings

Findings are a work queue, not a wall of text.

Filter by severity, confidence, check type or materiality; page through the list at your configured page size.
Open a finding to see the exact transactions, the rule applied and the supporting documents.
Resolve, accept or annotate each one — the decision and who made it is recorded in the audit log.

You end up with: A defensible workpaper trail for anything you were asked about later.

Step 07

Reports and future assumptions

The same ledger drives the historical statements and the forward view, so they can never disagree.

Read the P&L, cash flow and monthly trend for any date range and currency; click any line to drill into its entries.
Adjust the base, best and worst-case assumptions — revenue growth, expense growth, collection delay, one-off cash.
Compare the forecast against a backtest of how the same model would have performed on your real history.

You end up with: A six-month projection per scenario with closing cash and an auditable change log.

Step 08

Export, share and monitor

Everything on screen can leave the platform in a form an accountant will accept.

Export reports and audit workpapers to PDF or CSV, branded with your logo and including the evidence appendix.
Generate an investor-grade report with the forecast, confidence bands and the assumptions behind them.
Turn on CFO alerts to be notified in-app, by email or by push when cash, margin or exceptions cross a threshold.

You end up with: Reports you can send out, plus continuous monitoring between runs.

What the AI audit actually tests

Every check runs against the complete population for the period and is printed in the workpaper with the literal rule applied, so a reviewer can reproduce it by hand.

Duplicate payments

Same counterparty and amount inside your duplicate window.

Round-number entries

Suspiciously round amounts above materiality — usually estimates.

Statistical outliers

Amounts far from your normal spending pattern (median/MAD test).

Missing source document

Material entries with nothing attached to support them.

Uncategorised entries

Confirmed lines that cannot be mapped to an account.

Unreconciled ledger

Entries that do not balance or were never posted.

Period completeness

Months inside the range with no activity at all.

Off-cycle postings

Material amounts dated on a weekend.

Period cut-off

Material entries within days of the period boundary.

Direction conflict

Revenue that moves money out, or an expense that moves money in.

Benford first-digit test

First-digit distribution that deviates from natural populations.

Preparing your sheets before the first run

One file per account per period, covering the full period with no gaps.
Native PDF or CSV rather than a photo or a scan where you have the choice.
Statements that include the opening and closing balance lines.
Invoices and receipts for anything material you want matched to a ledger line.
A consistent base currency decided before the first run — it drives every report.

Reading the future assumptions

Forecasts start from the average of your last three recorded months, then apply the assumptions you set for each scenario: monthly revenue growth, monthly expense growth, a collection delay that shifts when cash actually lands, and any one-off cash movement you expect. Base, best case and worst case are projected side by side for six months, each ending in a closing cash position.

Every change to an assumption is diffed and logged — before value, after value, who changed it and when — and exports in the what-if audit trail, so a board or lender can see exactly which lever produced the number in front of them.