Regional leasing firm: from legacy systems to an AI-run operation

The context

The leasing firm operated on an old core, patched over many years. Generating regulatory reports relied on manual exports that took days per request. The monthly accounting close took weeks of the finance team’s time before results could be consolidated for the board.

The project was triggered by an external-audit observation on traceability of leasing approval decisions. Management decided it was time to migrate — without stopping origination or losing operating history.

How we approached the problem

A layered migration strategy with coexistence: the legacy core remained as the system of record throughout the transition while the new AI-centered stack absorbed processes by domain. Orchestration of Anthropic Claude, OpenAI GPT and Google Gemini for reconciliations, multi-model credit scoring and executive reporting.

First phase

Architectural audit and coexistence plan

Complete map of the legacy core: dependencies, integrations, critical batch processes. Event bus design for coexistence and cutover plan by domain.

Second phase

Automated reconciliations with agents

Bank reconciliation pipeline with document extraction and contextual matching. Manual reconciliation time dropped from the first weeks in production.

Third phase

Multi-model scoring and parameterizable credit policies

Replaceable scorecards without touching code, A/B testing in production and AI support for evaluating complex files.

Fourth phase

Accelerated accounting close and executive dashboards

Close automation: provisions, accruals, consolidated financial reports. Executive dashboards with LLM-driven explainability. The monthly close moved from weeks to days.

Fifth phase

Regulatory traceability and full go-live

Immutable log of approval decisions, signed electronic files, updated manuals. Total cutover with the legacy core in read-only mode. Start of the dedicated team.

Results

• Monthly accounting close in days instead of weeks.
• Bank reconciliations with far fewer manual hours from the finance team.
• Complete traceability of approval decisions, auditable, with evidence for external audit and CNBV.
• Greater origination capacity with the same commercial team thanks to the scoring engine.
• Operating history migrated to the new stack without information loss or commercial service interruption.

“We went from closing the month in weeks to closing it in days, and the external audit closed without traceability observations. The migration did not stop origination.”

— Finance Director, regional leasing firm · reference under NDA