Oracle says Tamilnad Mercantile Bank cut transfer decision time by 90%

Employee transfers that once required weeks of spreadsheet work and committee deliberation have become a same-day, AI-assisted process at Tamilnad Mercantile Bank, according to Oracle. In an October 14 case study, the technology supplier said the bank’s new transfer platform cut decision time by 90%, while keeping human review and approval in the workflow.

A transfer exercise spanning 628 branches

The bank’s annual transfer exercise covers more than 5,200 employees across 628 branches. It must match people to vacancies while balancing employee development and preferences, branch staffing needs, tenure requirements and Reserve Bank of India regulations.

Previously, HR staff manually assembled employee histories, performance information, qualifications, preferences and vacancy records from multiple sources. Oracle describes a process that consumed substantial senior HR and leadership time. Disconnected spreadsheets also made it harder to apply rules consistently, avoid data errors and demonstrate fairness in the resulting placements.

The replacement, called the TMB Transfer Matching Engine, brings those records into one application. It also includes dependents’ information and branch vacancies in its decision framework. The bank built it with Oracle APEX on OCI Autonomous AI Database, working with Kovaion Consulting on design and implementation.

Matching rules come before recommendations

At the center is a matching algorithm called Deferred Acceptance, also known as Gale–Shapley stable matching. Oracle says the engine evaluates more than 60 configurable business and regulatory policies before producing ranked transfer recommendations. Employee preferences sit alongside organizational priorities; they are not the only inputs used to recommend a placement.

• Eligibility and tenure: policies check whether an employee qualifies for a transfer and meets posting-duration requirements.
• Skills and career development: recommendations account for qualifications, certifications and career-exposure objectives.
• Placement constraints: the rule set includes employee preferences, organizational priorities, no-repeat-location policies and regulatory requirements.

AI assistance is one part of that arrangement. The application connects to OCI Generative AI, but Oracle does not explain exactly which recommendation tasks that service performs or separate its contribution from the matching algorithm and workflow changes. The reported speed improvement is a result attributed to the platform as a whole.

HR keeps the ability to change the outcome

HR teams can simulate transfer scenarios, inspect recommendations and make manual overrides before approvals. Dashboards and employee- and branch-level views give reviewers a shared application for examining proposed moves. Approval workflows vary by employee category, and committee approval remains part of the governance process.

Oracle describes ranked, explainable recommendations and complete audit trails as features supporting fairness and transparency. Role-based access controls limit access to sensitive employee and transfer data. These controls address another weakness of the old arrangement: information and decisions were spread across disconnected spreadsheets and reports.

The bank was already an Oracle Database and Oracle Fusion HCM Cloud customer. It chose APEX, Oracle’s low-code application platform, to build within that existing setup rather than introduce additional middleware or custom application frameworks. The resulting system connects workforce mobility with its existing HR data and Oracle cloud services.