The afternoon starts in an inbox and ends with line items re-keyed into a spreadsheet, while the purchase order and receiving report get chased through two other systems.
Accounting & Accounts Payable
The month doesn’t close on the clean invoices. It closes on the five exceptions.
For AP leaders, controllers, accounting firms, and CAS teams.
Start the day on the five invoices that actually need you — arithmetic already checked, the purchase order and the receipt already alongside, the disagreement already named.
invoices · purchase orders · receiving reports and receipts · statements · credit memos
Where the time goes
The hard part is finding the complete picture.
A clean invoice and a broken one look identical until someone re-adds the lines, checks quantities, and hunts down the receipt.
Every client packet arrives differently, so senior time goes to assembly, not review.
What reviewers see
One prepared review—not another packet to decode.
Vendor, invoice date, PO reference, subtotal, tax, total, and every line item — extracted, checked, and filed against the purchase order.
Three exceptions open. The clean lines are already off your desk.
In the example packet: the line items add up to 7,440 while the invoice says 8,060, the receiving report disagrees on quantity — five units against the six billed — and nobody signed the final receipt.
What Alembic handles
Move preparation out of the review queue.
Invoices arrive by upload, batch, or a configured inbox — header fields and every line item come back structured.
Purchase orders and receipts arrive as context. Configured checks — arithmetic, quantity, tax, required documents — run on every invoice and build an exception ledger.
Open the exception beside its source page, resolve or route it — only what you approved moves to Google Sheets, CSV, the REST API, or a signed webhook.
The accuracy question
We show you where we are unsure.
What this means for your team
Extraction is only worth trusting if it tells you when it is not sure. Where a number traces back to a page, that page opens beside it. Where it does not, the field says so instead of guessing. A quantity that disagrees with the receiving report stays a disagreement, both sources attached, until a person settles it. Nothing gets rounded off to make the queue look finished.
- The extracted value and its source page, one click apart, wherever a reliable match exists.
- Gaps and conflicts stay open. Approving a document never clears them behind your back.
- No score to interpret. Payment approval is a human act with a name on it.
Before you build a business case
Measure your own preparation work first
How to calculate your own baseline
Packets per month, minutes actually spent preparing each one, loaded hourly cost of the people doing it. That is your current cost of preparation, and it is usually larger than anyone expects.
Set it against the review effort you expect to keep, and put software and setup cost in the same view. Illustrative, not guaranteed: this is a capacity scenario, not a promised savings rate.
Where the line sits
What stays human
- Alembic does not match or pay invoices, and it does not approve payment.
- Source evidence is not available for every value or file; missing, uncited, or conflicting values remain review work.
- Purchase-order, receipt, and approval logic is set up for your workflow — it is not a prepackaged accounting module.
Accounting & Accounts Payable workflow questions
Does Alembic approve invoices for payment?
No. It prepares the fields, runs your checks, attaches the evidence, and names the exceptions. Accounting treatment and payment approval stay with your finance team.
Can purchase orders and receipts be compared?
Yes — you define the fields and the checks, and the supporting documents come in as context. Alembic shows you the difference; your team decides what it means.
Which destinations are supported?
CSV, REST API, signed webhooks, Google Sheets, and generic HTTPS destinations, subject to plan and setup. A named accounting system needs an adapter you configure — we do not imply native integrations we have not built.
How do we size the value without a made-up ROI?
Run a 20-invoice baseline on your own queue first: touch time, missing context, correction loops, exception age. Then model from your annual volume, active minutes, labor cost, and the review effort you expect to keep. We will not hand you a savings percentage we cannot stand behind.
Show us what is slowing your team down.
We’ll help you start with the documents, checks, and review points your team already uses.