Case study · Finance reconciliation and executive intelligence

A production AI system reduced a year of manual reconciliation from ~2,100 hours to ~5.

Across two companies, the system automatically reconciled 279 of 284 transactions. People handled five exceptions, while the CEO gained direct access to trusted financial information.

~2,095 hours returned to the team every year.

This was not a tool that helped people reconcile faster. Renaix redesigned the workflow so the system handles routine work, people handle uncertainty, and every result remains connected to its supporting evidence.

Owner-led European investment group. Anonymized by agreement.

The business case

What changed.

Before

~2,100

hours / year

Manual reconciliation work

2 people~60% of productive working time

After

~5

hours / year

Manual reconciliation work

5 exceptions~1 hour per exception

Result

~2,095 hours

returned every year

99.8%

less manual reconciliation effort

~1.2 FTE

capacity released

~90 hours

CEO time returned per year

The system did not make people reconcile faster. It changed the operating model: the system handles reconciliation; people handle the exceptions.

Manual reconciliation effort

99.8% reduction
Before
~2,100 hours
After
~5 hours

Transactions requiring human reconciliation

98.2% automatically matched
Total transactions
284
Human reconciliation required
5

Team capacity

The recurring workload effectively disappears
Previous reconciliation load
~1.2 FTE
Capacity released
~1.2 FTE

The executive problem

The real problem was not reconciliation. It was visibility.

The company operates across multiple entities, bank accounts, suppliers and financial periods.

For the CEO, this created a much bigger problem than administrative workload. Getting a reliable answer to a simple business question was difficult.

  • Which suppliers have we paid the most?
  • How much have we spent with a particular supplier across entities?
  • Which payments still have no supporting invoice?
  • Are there duplicate invoices or unusual payments?
  • What has one entity actually spent this year?
  • What information can we confidently give an investor or a bank?

Decisions are slower

A question that should take minutes can require emails, spreadsheets and manual investigation before the CEO can act.

The CEO depends on people to interpret the business

The executive does not have direct access to a trusted financial view. Important knowledge sits with the people who know where the data is, how it is structured and how to reconcile it.

Every new reporting request creates more work

Investor updates, bank financing, project reviews and management questions repeatedly require the same underlying financial information to be assembled again.

The finance team had the data. The CEO did not have the answers.

The objective was therefore bigger than automating reconciliation. It was to create a financial foundation the CEO could trust and use directly.

What Renaix built

One financial workflow across both entities

Renaix connected the company's documents, bank transactions and supporting financial records in one controlled workflow.

When the evidence supports a reliable match, the system reconciles the transaction automatically. When the evidence is incomplete or contradictory, it creates an exception for a person instead of guessing.

Every reconciled transaction remains connected to its supporting evidence, creating a trusted financial foundation for the CEO, the finance team and future reporting.

The operating model

Documents and bank data

Structured financial evidence

Transactions connected to supporting records

Does the evidence support a reliable match?

Yes · sufficient evidence

Automatically reconciled

279 transactions

No · uncertain

Human exception

5 transactions

Trusted financial foundation

Reconciled records connected to their evidence

  • CEO finance agent
  • Finance team support
  • Reporting
284

transactions processed

279

automatically reconciled · 98.2%

5

handed to people · 1.8%

Routine transactions are handled by the system. People retain control where the evidence requires judgement.

What happened to the five exceptions

When the evidence was not good enough, the system did not guess.

1.8%

Surfaced as exceptions

Five transactions could not be reconciled reliably from the available evidence and were routed for review.

  1. No guessed matches

    When the evidence was insufficient, the system stopped and created an exception instead of inventing certainty.

  2. Missing evidence made visible

    Each exception showed the documents, transaction data and unresolved gaps needed for review.

  3. The accountant made the decision

    The system prepared the case, but the finance team determined the final reconciliation.

  4. Evidence stayed attached

    Once resolved, the record retained the supporting documents and decision trail behind the outcome.

The system is trustworthy not because it automates 98.2% of the work, but because it makes the remaining 1.8% explicit, controlled and auditable.

The finance agent

One finance agent. Two jobs.

The reconciled data also powers a finance agent used by both the CEO and the accounting team.

Trusted financial foundation

One finance agent

Answers

CEO

  • Gets trusted answers
  • Questions reconciled records
  • Uses information to decide

Exceptions

Finance team

  • Investigates uncertain cases
  • Reviews transactions and evidence
  • Retains the final reconciliation decision

Use case 01

The CEO can ask the business directly

Instead of asking someone to assemble the information, the CEO can question the reconciled financial data directly.

Representative Finance Agent exchange based on the production system
  • CEO · Sunday 15:04
    What is our consolidated liquidity across both entities?
  • Finance Agent
    It's €1,284,506.32.
    Holding entity: €961,210.11 as of 31 Dec.
    Operating entity: €323,296.21 as of 30 Dec.
    All Reconciled against bank statements.
  • CEO
    Anything pending I should know about?
  • Finance Agent
    Three transactions await review. The largest is €18,435.20, unmatched for six days. Every figure links to its source document.
  • CEO
    Who acts on that?
  • Finance Agent
    It is queued for your finance lead with the evidence attached. Nothing has been booked without human approval.

The CEO can ask about:

  • Supplier spending
  • Payments across entities
  • Missing documentation
  • Unusual transactions
  • Spending over time
  • Other information contained in the reconciled data

Every answer remains connected to the available evidence.

Before

Ask someone to find, reconcile and explain the answer.

Now

Ask the financial system directly.

Use case 02

The finance team investigates the exceptions

When the system cannot establish a reliable match, the agent helps the finance team understand why.

It brings together:

  • The bank transaction
  • The available documents
  • Related financial records
  • The evidence supporting possible matches
  • The information that is missing or contradictory

The agent prepares the investigation. The accountant reviews the evidence and makes the final decision.

Measurement and scope

What was measured, and what was estimated.

Observed system results

284 bank transactions, one full financial year

  • 279 automatically matched
  • 5 required human reconciliation
669

documents processed

284

bank transactions processed

279 of 284

automatically matched

98.2%

automated match rate

5

unmatched exceptions

1.8%

exception rate

2

duplicate-invoice suspects identified

2

foreign-invoice gaps identified

2 entities · one fiscal year

measured scope

Estimated people impact

~2,100 → ~5

hours of manual reconciliation per year

~2,095 hours/year

team capacity returned

~99.8%

reduction in manual reconciliation effort

~1.2 FTE

capacity released

~90 hours/year

CEO time returned

System-performance figures are measured from the production dataset. People-impact figures are estimates based on the team’s previous workload and the observed exception volume.

Beyond one workflow

Proven on a small part of a much larger estate

  • 2 entities × 1 financial year2 entities · one fiscal year
  • 17 entities × multiple financial periods17 entities · 2024 · 2025 · 2026 year-to-date

A single year across all 17 entities represents 8.5× the entity scope of the measured case study.

Different entities can carry different transaction volumes and complexity, so we do not present this multiplier as realized savings.

But the operating principle remains the same: once reconciliation becomes a system capability rather than a recurring human task, the benefit grows with the amount of finance work brought into the system.

From back-office work to finance infrastructure

Before

  • People manually reconcile the data.
  • The CEO asks people for answers.
  • New reporting requests create more manual work.

Now

  • Transactions and documents are reconciled automatically.
  • People work only on exceptions.
  • The CEO can question the reconciled data directly through the agent.
  • The finance team uses the same agent to investigate remaining exceptions.

Next

  • Investor reporting is generated from the same financial foundation.
  • Bank reporting supports financing for future investments.
  • More entities and financial periods are brought into the system.

The benefit is therefore not only the ~2,095 hours of annual reconciliation effort already removed. It is the creation of financial infrastructure that can be reused across the group.

Bring your critical workflow.

If improving one workflow would materially change the economics, speed or reliability of your operation, we should look at it.

Schedule a fit conversation