Every business has a system of record. Almost none has a system of resolution.
Exception handling across orders, invoices, payments and deliveries, alongside the ERP you already run. What that means for a mid-market back office, and why the AI should never post to your ledger.
Record what fits. Resolve what doesn’t.
01The idea in full
A system of resolution is the layer that sorts what doesn’t fit the system of record, before it is recorded and after, so the ledger only ever takes what fits tolerance.
Every business has a system of record, the ERP, and it accepts a transaction once it fits within the tolerances you have set: an invoice matched to an order and a receipt, a payment matched to the invoices it settles. Reality doesn’t arrive like that. It arrives as PDFs in an inbox, remittances that don’t add up, deliveries that differ from what was ordered. So in most mid-market manufacturers and distributors, the back office’s real job has quietly become making reality fit the ledger, by hand, and the part that doesn’t fit is where their week goes.
Almost every business has a system of record. Almost none has a system of resolution, the place where everything that doesn’t fit gets sorted, before it can be recorded, and afterwards, when a recorded invoice is disputed or a payment arrives that doesn’t tie out.
Today that system is a person, an inbox and a spreadsheet, and it is why back-office headcount rises with revenue.
Apridata built Opnua to be that system. Rules handle what fits, because a rule can be audited and a rule never has a bad day. An AI agent works what doesn’t: it reads the correspondence, contacts the supplier or the customer, recommends the resolution and logs every step. It never posts to the ledger: rules post what fits, and a named person approves what it recommends. We don’t replace your ledger.
We do this because the most skilled people in a finance function shouldn’t be its integration layer. They should be doing the analysis the board keeps asking for.
Rules record. AI resolves. People approve.
The resolution tax
The resolution tax is the avoidable cost of investigating and clearing exceptions across systems, documents and people; it is why back-office headcount rises with revenue.
Skilled people’s hours spent copying, checking and chasing, plus the cash sitting uncollected while they do. Its consequences, capacity consumed, cash collected late and a close that lands later than it should, are related, and are counted separately.
How many people would your back office need if every order, confirmation, delivery, invoice and payment arrived already agreeing with the one before it?
The gap between the answer and today’s headcount is the capacity resolution is consuming: a growth question, not a redundancy plan.
Three beliefs
Rules before AI
Rules handle the routine majority, because a rule can be audited. AI belongs on the judgment work: the exceptions and the investigations. Start with measurable problems, not another technology project.
Back-office capacity shouldn’t scale with transaction volume
Headcount rising with revenue is the sign that resolution is being carried by hand. Transactions are absorbed by rules and AI; the team grows in analysis, not processing.
From producing numbers to acting on them
For the controller this is control: a ledger trusted sooner. For the director it is commercial: capacity moved from producing information to influencing decisions.
What this keeps out of your ledger
Configured rules post eligible transactions. The agent investigates what doesn’t fit and recommends a resolution. Anything it proposes is written back only after a named person approves it. Every step is logged, so the audit trail shows what was read, which policy applied, what was recommended and who said yes. The AI has no write access to your ledger, and that is a design decision, not a setting.
Post what fits
Deterministic, auditable, and the same answer every time. If it can be a rule, it is a rule.
Investigates and recommends
Reads the correspondence, works out what happened, proposes a resolution with its reasoning and evidence.
Approves
Accepts, changes or rejects it. Only then does it post. Every step, by whoever or whatever took it, is logged.
A path from the AI to the ledger
The agent has no write access to your ERP. That is a design decision, not a setting.
The five places resolution happens by hand
Each carries the question that turns agreement into a number.
- 01 · Intake
Documents arrive as email and PDF
And get keyed into the ledger by hand.
How many documents did someone type into your system last month, and what else could that person have done?
- 02 · Matching
Exceptions route through inboxes
The three-way match clears the clean cases. The rest live in email, phone calls and memory.
Where do your mismatched invoices actually live while they wait, and who knows they’re there?
- 03 · Cash
Remittances don’t tie to the bank
Payments arrive short or combined. Unallocated cash sits on the account for weeks.
How much cash is sitting unallocated right now, and how long does a remittance that doesn’t tie out take to clear?
- 04 · Collections
The chase lives in one head
Or one spreadsheet. Disputes stall because the paperwork is in three places.
If your credit controller was out for a month, what would happen to your debtor days?
- 05 · Close
Spreadsheets bridge the systems
The close lands on the day it lands because of how many bridges there are.
What day does your close land, and which spreadsheet is the reason it isn’t earlier?