Most financial operations work exists to answer two questions no single system can: what actually happened, and was it what was supposed to happen?
The first question shows up in reconciliation. The processor file says one thing, the bank statement says another and the ledger says a third. Someone has to work out what actually happened before the day closes.
The second shows up everywhere else. A settlement due Tuesday has not arrived by Thursday. The fee charged is not the fee in the agreement. A policy required an escalation, and nobody escalated.
None of that means a system failed. The processor, the bank and the ledger may each have recorded their part correctly. Each one knows only its part, and the work starts in the space between them.
Institutions fill the space between systems with people
Every institution calls the work something different: reconciliation, exceptions, controls, payment operations, oversight. The job is largely the same. Take records from several systems, work out why they disagree, decide which rule or agreement applies and decide what happens next.
That work rarely shows up on the budget as software. It shows up as people. As transaction volume grows, the team grows. Every new processor, sponsor bank, rail or corridor adds another set of handoffs someone has to understand. No institution set out to build a layer of people whose job is to make systems agree. Each one accumulated it.
More software made each system better and left the work between them untouched
Most financial software was built to make one system or one function better at its own job. The ledger records, the processor processes, the compliance system runs its controls and the case manager manages the case. All of it can work exactly as designed while the operator still has to answer what happened across all of them, and whether it matched what should have happened.
That is why the work gets mistaken for a data problem: put everything in one warehouse, build another dashboard, give the team a better view. That helps, and it leaves the hard part. The operator needs more than proof that two records differ. They need to know which one matters and why.
That takes something most operational systems do not hold: the expectation. The fee schedule, the settlement term, the partner agreement, the network rule and the internal policy. A reconciliation system can tell you two values do not match. On its own it cannot tell you which value the governing agreement says is correct. That judgment has lived in people's heads, documents and operating procedures, so the exception goes to a person.
Software can now read what should have happened and compare it with what did
Both sides of the problem can now sit in one place. On one side, the records of what actually happened across the financial stack. On the other, the contracts, policies and rules that say what should have happened. Turn those requirements into operating rules, keep each one tied to the document it came from, and compare them continuously against the activity underneath.
The operator no longer starts with four systems and an agreement PDF. They start with what happened, what should have happened, where the two diverged and the evidence for each.
Cordant does that comparison across the systems you already run
Cordant is the command center for modern financial infrastructure. It replaces no processor, ledger, bank or case manager. It reads the records those systems already produce, connects them to the agreements and policies that govern the operation, and finds where actual behavior diverges from expected behavior.
When something goes wrong, the exception arrives with the cause and the evidence attached, in the workflow where the team already works. The ledger stays the ledger, the case stays in the case manager and the operator still decides what to do. What goes away is much of the investigation it took just to understand the problem. The person who spent the afternoon rebuilding the answer spends it deciding what the institution should do.
Count the hours your team spends between systems
Open last week's exception queue, the reconciliation workbook, or the notes from the standing meeting where everyone brings the transactions nobody can explain. Count the hours spent answering the two questions: what actually happened, and was it what should have happened? That number is the work your existing systems leave behind.
It grows as institutions add rails, counterparties and agents. An agent that crosses systems hits the same problem a person does. It needs to know what happened and what was supposed to happen, and until both answers sit in one place, automating the work only gets it to the unanswered question sooner.
