Running for early-access clients Every line reviewed by an accountant

Bookkeeping where the ledger codes itself and a person signs it off

We have built AI into the bookkeeping we already do. Transactions are categorised as they land, anomalies are flagged the same day, cash is forecast 90 days out and your monthly reports arrive with an explanation β€” all checked by a qualified TOA accountant before you see them.

95%+of transactions coded without a human touching them
3Γ—faster month-end close for early-access clients
100%of AI output reviewed before it reaches you
Chase Business Checking Β· bank feed AI coding 0/8
02 Feb Stripe payout β€” 118 orders +$9,412.60 Sales revenue
02 Feb Google Workspace -$144.00 Software subscriptions
03 Feb Northgate Supply Co. -$1,286.40 Cost of goods sold
03 Feb ADP payroll run -$18,204.11 Payroll & benefits
04 Feb Adobe Creative Cloud -$59.98 Duplicate β€” review
04 Feb Chase card payment -$3,000.00 Credit card transfer
05 Feb Meta Ads -$2,110.00 Marketing
05 Feb Wire from Harbor Builders +$14,000.00 Accounts receivable
1 item routed to your accountant Β· 7 coded from prior patterns Reviewed by your TOA accountant Β· ready for close

Reads from the systems you already run

How the month runs

AI does the volume. Your accountant does the judgement.

The close is a sequence. The two lanes show who does what on which day β€” AI takes the repetitive early work so human review happens sooner and on fewer items.

Day 1 Β· bank feeds land
Day 2–3 Β· coding
Day 4–5 Β· review
Day 6–7 Β· delivery
AI
Ingest & normalisePulls transactions from every connected account and system into one ledger.
Code & matchCategorises against your chart of accounts, matches payouts, invoices and POs.
Flag & forecastSurfaces duplicates, spikes and unmatched items; refreshes the 90-day cash forecast.
Draft commentaryWrites the first draft of what changed and why, with the numbers to back it.
Accountant
Nothing to do yet β€” the feeds are still arriving.
Judgement callsCodes new vendors, owner draws, mixed-use items and anything the AI was unsure about.
Reconcile & challengeTies every account, checks each flag and decides what actually needs your attention.
Sign off & sendEdits the commentary, approves the dashboard and closes the month.
What the AI actually does

Six things that used to take hours

Each capability is paired with an example of what lands in your inbox or on your dashboard β€” real formats, illustrative numbers.

Book a demo with your data

Transaction categorisation

A model trained on your chart of accounts and coding history assigns every incoming transaction. Confidence below threshold routes the item to your accountant with the AI's reasoning attached.

QuickBooks Online Β· Xero Β· Sage Β· NetSuite Β· 2–3 month calibration to reach 95%+
Review queue item"Northgate Supply Co. β€” $1,286.40. No prior history. Nearest match: Cost of goods sold (71%). Vendor name suggests materials supplier."Needs judgement

Anomaly and duplicate detection

Every entry is checked against your normal patterns β€” vendor, amount, timing and frequency. Duplicates, expense spikes, unmatched deposits and invoices that disagree with their PO are flagged the day they appear.

Runs nightly Β· alerts ranked by dollar impact
Alert Β· 4 FebAdobe Creative Cloud charged twice within 3 minutes on card ending 4417 β€” $59.98 each. Refund request drafted.Duplicate payment

Cash flow forecasting

A rolling 30/60/90-day projection built from your history, open invoices with their real payment behaviour, scheduled payables and payroll. Re-run each night as the ledger changes.

Seasonality-aware Β· scenario toggles for large receipts and hires
Forecast note Β· FebruaryCash dips to $118,200 around 18 Feb as payroll and quarterly insurance land together. Collecting Client A ($8,400) by the 14th removes the dip.Watch

Payout, invoice and PO matching

Marketplace and processor settlements are tied back to orders, fees, refunds and reserves. Supplier invoices are three-way matched to POs and receipts. What does not tie is surfaced, not buried.

Shopify Β· Amazon Β· Stripe Β· A2X Β· Buildertrend Β· NetSuite
Settlement Β· Amazon #A-882131,204 orders Β· $48,222 gross Β· $6,904 fees Β· $41,318 net. Deposit received 3 Feb matches to the cent.Tied

Questions in plain English

Ask your own ledger a question and get an answer with the underlying figures. It only ever reads your books β€” nothing is used to train public models.

Available in the dashboard and by email
You asked"Which clients are more than 30 days overdue?"

AnswerThree: Larkin LLC $8,400 (44 days), Pine Dental $5,800 (37 days), Voss Media $3,700 (32 days). Follow-ups sent 1 Feb.From live AR ledger

Monthly commentary and reports

Your P&L, balance sheet and cash flow arrive with a short narrative on what changed, why, and what to do about it. AI writes the draft; your accountant edits and signs it.

Delivered by the 7th Β· also on your dashboard
January commentary (excerpt)Revenue up 12.4% on three new construction clients. Gross margin held at 64.2%. Marketing +47% with no attribution yet β€” request ROI breakdown before approving February.Signed off by your accountant
Three ways to work with us

Pick the layer you need

AI is already inside every bookkeeping package we sell. If you need more than that, the two engagements on the right go deeper.

AI inside your bookkeeping

Included in every small business, e-commerce, construction and B2B engagement.

  • Overnight transaction coding
  • Anomaly and duplicate alerts
  • 90-day cash forecast
  • Monthly commentary
See bookkeeping packages

AI dashboard and analytics

One live page replacing the five to seven reports your team produces by hand.

  • Real-time KPIs from every connected system
  • Forecast, AR aging and alert feed
  • Ask your ledger questions
  • White-label option for firms
Explore the dashboard

Business process optimisation

For finance teams whose systems do not talk to each other yet.

  • ERP, CRM and payroll integration
  • AP/AR automation
  • Data migration and consolidation
  • Workflow redesign and SOPs
See how a process review works
Questions we get asked

Honest answers about AI and your books

Will AI replace my accountant?
No. AI is good at pattern recognition, repetitive categorisation and watching large volumes of data around the clock. It cannot exercise professional judgement, interpret an ambiguous transaction, advise on tax or understand the context of your business. We use it for the routine work and route everything that needs judgement to a qualified accountant. Nothing reaches you unreviewed.
Is my financial data safe when AI processes it?
Yes. We connect through read-only APIs (never passwords), process data in isolated encrypted environments, and every team member with access is under NDA. Your data is never used to train public AI models. The security posture is the same as sharing data with any qualified accounting firm β€” which is what we are.
How accurate is the categorisation, really?
Above 95% after a two-to-three-month calibration on your chart of accounts. The first month needs more human review while the model learns your patterns. Any transaction the AI is unsure about is routed to your accountant before the books are finalised.
What software does it connect to?
QuickBooks Online, Xero, Sage, Oracle NetSuite, Microsoft Dynamics 365, SAP Business One, Shopify, Amazon Seller Central, Stripe, A2X, Buildertrend and most major payroll platforms. If your stack is not listed, ask β€” integrations are added regularly.
What does it cost compared with standard bookkeeping?
AI is now included in our standard bookkeeping packages at no extra charge. The dashboard and process-optimisation engagements are priced separately, and we will always show you the cost against the hours it removes before you commit.

See it running on your own numbers

We are onboarding a limited number of early-access clients. Book a free 30-minute demo and we will connect the dashboard to your QuickBooks or Xero before the call.