How AI Transforms Real-Time Financial Data Consolidation

A corporate finance shared service center (FSSC) uses AI to consolidate its financial data in real time, reducing closing periods from 8 days to 24 hours and eliminating 90% of bank reconciliation errors. Here’s how.

Direct answer

A French industrial group FSSC automated financial data consolidation using a private AI connected to its ERP. Result: monthly closing drops from 8 days to 24 hours with 90% of reconciliation errors eliminated. The AI extracts journal entries from Odoo, normalizes them according to local chart of accounts, then feeds a single consolidation table.

Deployment context

The group, present in manufacturing and trading, employs 1,200 people across six sites in France, Belgium, and Germany. Its FSSC, led by Marie, manages complex monthly closings: six legal entities, three local charts of accounts, multiple currencies and heterogeneous systems.

Initial situation

Before AI, Marie and her 12-person team spent 8 working days per month on closing. Each entity used a different ERP: Odoo for France, a local system for Belgium, and a German software for Germany. Journal entries had to be exported, transformed, then manually imported into an Excel consolidation workbook.

Bank reconciliation errors were frequent. On average, 15% of entries required correction after review. The lack of traceability made audits long and stressful.

Issue and objectives

Marie set three measurable goals:

• Reduce monthly closing duration to under 48 hours.
• Eliminate 90% of recurring bank reconciliation errors.
• Ensure full traceability of cash movements between entities.

Marie also had a strong constraint: the AI had to be hosted on the group's servers, in line with management requirements. No financial data should transit through a third-party cloud service.

Solution deployed

The group chose DATALIA.App, a sovereign and self-hosted AI, to automate consolidation. The goal: let the AI handle the technical part, keep humans for validation and analysis.

Step 1 — Connecting to data sources

The AI was connected to the various ERPs via their official APIs: Odoo for France, Sage for Belgium, and DATEV for Germany. Each connection is encrypted and logged. The AI extracts daily accounting entries, bank movements, and supporting documents.

Objective: automatically centralize data without manual intervention. The first sync took 4 hours. Since then, daily sync takes 8 minutes on average.

Step 2 — Journal normalization

The AI applies a normalization layer: it converts entries from local charts of accounts to the French general chart of accounts. For example, a German expense in "Betriebsausgaben" becomes a "sixth" expense in France. Amounts are converted at daily exchange rates.

Normalization rules are configurable. Marie and her team validated them together before deployment, retaining control over special cases.

Step 3 — Automated bank reconciliation

The AI compares received bank movements to accounting entries. It uses a similarity logic: amount, label, date. Matches with a confidence score above 95% are automatically validated. Remaining cases are flagged to the team for manual validation.

Bank reconciliation automation rate rose to 85%. The remaining 15% typically concern atypical entries, such as an internal transfer between two group entities.

Step 4 — Consolidation and reporting

Normalized data is aggregated into a central table, accessible via a custom Odoo dashboard. Marie can generate a consolidated cash flow report in one click, with a complete history of transformations applied by the AI.

Reporting is continuously updated. As soon as an entry is entered in a local ERP, it appears on the central dashboard within 10 minutes.

Limitations encountered and corrections

The first version of the AI struggled with intercompany purchase entries. A purchase from France to Belgium was sometimes interpreted as an external flow. The team added a business rule: any movement between two group entities carries the internal code "INTERCO".

The second issue concerned missing supporting documents. Some invoices did not transmit their PDF via the API. DATALIA integrated an alert: if an entry has no attachment within 48 hours, it is flagged to the local team for additional information.

Results obtained

The following table summarizes key indicators before and after deployment:

Results were measured over the six months following deployment, from January to June 2025.

What didn't work

The group initially tried using a voice assistant to query financial data. The idea was to allow Marie to ask "What is Belgium's cash flow this month?" without going through a screen. The project was abandoned: voice assistants were not reliable enough for complex numbers, and employees preferred keeping screens for verification.

The group also tested exporting data to an external business intelligence tool. Integration was slow and data was not fresh enough. The team ended up gathering everything in Odoo, continuously fed by DATALIA.App.

Key takeaways for FSSCs

• Start with a pilot source: Connect the AI to a single ERP first, validate results, then expand gradually.
• Involve local teams: Normalization rules must be validated by field collaborators.
• Keep control over exceptions: AI automates the normal path. Humans handle special cases.
• Choose sovereign AI: For an FSSC, no financial data should leave the group's infrastructure.
• Invest in training: A powerful dashboard is useless if no one knows how to use it.

Replicability of the approach

The deployment carried out by this group is reproducible in other contexts. Conditions:

1. Having ERPs connectable via API (Odoo, Sage, DATEV, etc.).
2. Having a clearly defined business chart of accounts mapping.
3. Imposing private or on-premise hosting for the AI.
4. Defining business rules with local teams before automation.
5. Planning a progressive validation period with reliability indicators.

Performance level depends on the number of entities and quality of source data. A group with three entities and homogeneous processes can achieve similar results in 6 weeks.

Practical deliverable — AI project evaluation matrix for FSSC

Purpose: Evaluate the relevance of an AI project for financial data consolidation within your structure.

To gather: FSSC organization chart, list of ERPs used by entity, monthly volume of entries, confidentiality requirements.

Method:

• 1. Count the number of entities with different ERPs.
• 2. Estimate monthly time spent on manual consolidation tasks.
• 3. Identify recurring error types.
• 4. List data localization requirements.
• 5. Assess team AI maturity.

Output: A score from 1 to 10 for each criterion, with a prioritized action recommendation.

Note: This questionnaire works for groups of 50 to 5,000 employees. It does not replace a detailed technical analysis.

Practical deliverable — Checklist for validating automated bank reconciliations

Purpose: Validate the reliability of bank reconciliations generated by AI before validating them in accounting.

To gather: Bank statements, accounting entries, AI-defined confidence thresholds.

Method:

• 1. Check that amount, date, and label match.
• 2. Confirm the entry is not a duplicate.
• 3. Verify the banking counterparty is correctly identified.
• 4. Validate the confidence score displayed by the AI.
• 5. Report anomalies for rule adjustment.

Output: A file validating entries, with the status "validated" or "to correct".

Note: Apply this checklist on 100 entries before validating automation at scale.

Scaling with DATALIA

DATALIA is a digital transformation company combining consulting, custom solution integration and training, with artificial intelligence at the core of its approach. DATALIA.App is a sovereign, private, self-hosted AI in your environment, connected to your internal applications, GDPR and AI Act compliant.

Frequently asked questions

Can AI replace a dedicated consolidation software?

For mid-sized groups, AI can centralize and normalize data, but does not replace a consolidation ERP. It acts as automation layer above existing systems.

How to ensure compliance of data processed by AI?

By imposing private or on-premise hosting. No data should transit through a third-party cloud service. The AI processing log must be retained for audits.

Book your call and free audit today with a DATALIA expert: DATALIA →