Integrating AI agents into banking operations
In a bank, the real question is where the AI runs and who validates. Crédit Agricole du Maroc has chosen open source models installed on its own infrastructure, and local solutions for its banking files, reporting and project management.
The challenge
Legal files arrive in free format, often scanned, and each document must be read, cross-referenced and then re-entered into the central system before a document can be produced. Time-consuming work, prone to input errors, and difficult to scale up. The additional constraint comes from the sector: no data can leave the bank, no remote proprietary model can be called, and no document can be sent without a lawyer having validated it.
7 AI use cases developed for banking
AI-assisted releases
Prepares release certificates from file documents, with OCR extraction and consistency checks.
- Extract information using OCR and LLM
- Check data concordance between parts
- Generate the certificate for validation by the lawyer
Deliverable: Release certificate prepared, ready to verify.
Processing of guarantee files
Classifies the scanned documents, extracts the data and prepares their synchronization with the bank's IS.
- Classify documents and extract their data using OCR and LLM
- Structure metadata in the expected format
- Synchronize folders via SFTP
Deliverable: Classified file and structured metadata for the IS.
Extraction of balance sheets for credit analysis
Transforms PDF balance sheets into Asset, Liability and CPC data that can be used in Excel or JSON.
- Read and extract financial tables
- Clean and structure the Asset, Liability and CPC sections
- Export data for credit analysts
Deliverable: Balance sheets structured in Excel or JSON for credit analysis.
Point of sale quality reporting
Centralizes network evaluation campaigns and writes reports with an LLM.
- Centralize the collection of reviews
- Consolidate results by campaign
- Generate quality reports
Deliverable: Consistent evaluation reports, campaign by campaign.
HR collection and reporting
Organizes HR collection campaigns, validations and formula creation with AI.
- Collect data by campaign
- Have information validated by managers
- Assist in the creation of analysis formulas
Deliverable: Validated HR data and formulas for reporting.
Project planning assistant
Generates a structured project plan and tracks tasks and deadlines with alerts.
- Break down the project into tasks
- Organise deadlines and follow-up
- Report deadlines to monitor
Deliverable: Project plan, task list and tracking alerts.
Structuring recovery files
Extracts data from files, checks it with the teams then exports it to the IS format.
- Extract and classify parts by OCR and LLM
- Have the data checked and corrected by a colleague
- Map fields and export to IS
Deliverable: Files structured, validated and ready to integrate.
The results
On the release flow, process productivity was multiplied by four and human errors reduced by 95%. For legal files, re-entry in the central system has disappeared and updating becomes instantaneous. Each new agent is added on the same infrastructure and under the same control framework, which makes the next one faster to put into service than the previous one.
Define the framework for deploying AI in banking
Prepare the integration of generative AI into banking professions. The mission links uses to their security, architectural and organisational prerequisites.
Identification of uses with the teams, then formalization of inputs, outputs and expected benefits.
Comparison of solutions and hosting methods with regard to banking data.
Central coordination, AI Champions relay and steering committee to arbitrate initiatives.
Development phases, training, communication and performance monitoring.
Framed uses
Internal knowledge base, customer support, credit analysis, business intelligence and legal assistance.
Use case sheets, architecture options, a governance plan and an implementation schedule. Demonstrations help prepare for the development stages.
Let’s discuss your project
Let’s talk about your AI project
Tell us what you need. Our team will contact you to clarify your objectives and next steps.
