EXPERIMENT 01PROTOTYPE

OmniRecon

Governed reconciliation for systems that disagree

Can software reconcile conflicting records across systems while keeping automated decisions bounded, auditable and subject to policy?
Finance and operations teams reconciling records across billing, orders, fulfilment, supplier or settlement systems.Open OmniRecon demo

THE PROBLEM

What is actually going wrong

SaaS revenue reconciliation is the first use case, not the whole problem. Any business with several systems describing the same activity eventually has to decide which record is right and what to do when they disagree. The same pattern appears in orders and fulfilment, supplier invoices and goods received, and payments and settlement.

Today that work often falls to experienced analysts. They pull exports, map identities, rebuild the comparison, investigate exceptions, make a correction and then chase confirmation that it landed. The expensive part is not finding a difference. It is the hours of skilled work needed to establish what happened and repair it safely.

When that work is delayed, the cost compounds. An incorrect charge can become a refund, a support case and a trust problem. A missed charge becomes lost revenue. In either direction, the explanation is often buried in a spreadsheet or Slack thread, making the next investigation take just as long.

WHAT I DID ABOUT IT

I built

A reconciliation platform that compares source records automatically, settles straightforward matches and creates a review case only where systems still disagree.

In the implemented prototype, the agent helps triage unresolved cases. Investigation and corrective proposals are designed boundaries, not completed autonomous workflows. A separate policy engine decides whether any future action could be automatic, needs human approval or is denied.

A correction is checked on the next ingestion pass rather than assumed to have worked. The case and event log keep the evidence, decision and outcome together.

ENGINEERING NOTESDeep dive

TRY IT YOURSELF

Demo

The application runs separately. It opens in a new tab.

Open OmniRecon demo

BEFORE AND AFTER

What changed

Before

Illustrative: 20 to 50 hours/month
  1. 01Pull the usage export
  2. 02Pull the billing export
  3. 03Reconcile identities by hand
  4. 04Aggregate usage to the invoice grain
  5. 05Build the comparison sheet
  6. 06Triage flagged rows
  7. 07Investigate each survivor
  8. 08Decide and act

After

Pilot measure: analyst hours, case age and unresolved value
  1. 01Ingest source records as facts
  2. 02Resolve entities and match deterministically
  3. 03Settle correspondences within tolerance
  4. 04Open cases for unresolved residue
  5. 05Triage unresolved cases for reviewer attention
  6. 06Verify any tested correction on the next ingestion

SOURCE RECORDS

Exports and mapping tabsProvenance-stamped facts

MATCHING

Spreadsheet lookupsDeterministic correspondence rules

UNRESOLVED WORK

Flagged rowsCases with evidence and policy state

CORRECTION FOLLOW-UP

Usually uncheckedSelf-reconciliation on next ingestion

ARCHITECTURE

How does it work?

RECONCILIATION FLOW

  1. INPUT

    Source systems

    Metering, billing and rate records.

  2. BOUNDARY

    Scoped connectors

    Gated reads and writes.

  3. STORE

    Fact store

    Immutable provenance.

  4. RULES

    Matcher

    Rules and tolerances.

  5. RESIDUE

    Case workflow

    Investigate unresolved residue.

  6. GOVERNANCE

    Policy and verification

    Authorise, then verify.

Only unresolved residue reaches the case workflow. Matching, authorisation and verification stay deterministic.

The split is deliberate: models help with ambiguity, while matching, authorisation and arithmetic remain outside the model. The engineering notes include five interactive diagrams of the implemented lifecycle.

BUSINESS IMPACT

What could this mean for a real business?

  • POTENTIAL

    The approach could return skilled analyst time to exceptions that genuinely need judgement, instead of repeatedly assembling evidence and checking straightforward records by hand.

  • POTENTIAL

    Resolving and verifying a discrepancy before it reaches customers or the close process could reduce the cost of refunds, support work, write-offs and repeated investigation.

Observed means measured in a real engagement. Estimated is reasoned from the work but not measured. Potential is what the approach makes possible. Nothing here is dressed up as more than it is.

CODE

See the implementation

ARCHITECTURE ONLY

The source code remains private while OmniRecon is in progress. Request access to the reviewed architecture and decision records.

Request architecture access

SHARE AND ENJOY

Share and enjoy

Tell me what's going on. I'll come back within a business day, and if I'm not the right person for it I'll say so.

What are you after?

Pick as many as apply.

50 people
1200+

First step is a free half-hour call. No pitch deck, no obligation. Prefer plain email? Reach me directly at hello@magratheastudio.com.