The Union Internationale de Banques (UIB) is a Tunisian bank founded nearly 60 years ago, a subsidiary of the Société Générale Group. With 145 branches and approximately 900,000 clients, UIB is a key player in the Tunisian banking landscape. Like any long-established universal bank, UIB must contend with an extensive technical legacy — dozens of systems, heterogeneous data, and R and Python code accumulated over the years — while also responding to a highly competitive local banking market and increasingly strict regulatory requirements.

UIB's challenge came down to three words: volume, heterogeneity, acceleration.

‍

On one hand, dozens of systems accumulated over 60 years, fragmented data sources, shadow IT Excel and VBA in disarray, R and Python code to maintain, and the classic difficulty for large banks to leverage their contextual data at scale.‍On the other hand, a time to market to reduce to remain competitive, and a strong ambition: to move AI from discourse to production, within a demanding banking regulatory framework (compliance, auditability, internal controls aligned with Société Générale Group standards).‍A conviction structured the approach from the outset: AI could not simply be added on top of existing systems. To transition to AI, it was first necessary to properly structure the endeavor — lay a foundation, ensure data reliability, and industrialize a sustainable approach.

‍

Rather than multiplying isolated PoCs, UIB chose an atypical approach for an AI project: starting with the foundations.

‍

With Prime Analytics and on the Alteryx platform, UIB deployed a Smart Data Hub — a unified analytical foundation that centralizes data from all systems in one place.

‍

At the heart of this foundation:

• Alteryx workflows for multi-source data collection (structured and unstructured),

• automated quality controls and data cleansing,

• an operational governance discipline designed for the banking context.

‍

Based on this, the team deployed an automation factory: an industrial mode for producing business use cases, founded on shared rules and a unified technical stack. Among the use cases already deployed or in progress:

‍

• Behavioral customer segmentation — unsupervised AI clustering on the Retail portfolio (K-Means algorithms), to guide marketing and sales actions.

‍

• Internal Corporate Rating System — a probability of default model, complementary to existing Société Générale Group models, adapted to the specificities of Tunisian regulations.

‍

• Internal IFRS Rating — extension of the rating system to IFRS requirements, in an area particularly scrutinized by regulatory authorities.

‍

• Model Monitoring — Alteryx tools to track scores, detect deviations, and ensure the absence of AI hallucinations in critical pipelines.

‍

A rigorous evaluation discipline — both statistical AND business-oriented — was implemented from the outset, with continuous monitoring to ensure models remain reliable over time.

‍

UIB's trajectory demonstrates the value of a platform-first approach: 5 use cases deployed in the first year, with over a dozen underway for the next. Beyond the number of cases, the benefits are structural.

‍

• Significantly reduced processing time — A significant portion of the time spent on data processing has simply vanished, freeing up teams for value-added tasks.

‍

• FTEs saved — Particularly valuable in the current banking landscape, where operational agility has become a competitive imperative.

‍

• Enhanced compliance and auditability — Essential in IFRS and credit risk domains, closely monitored by regulators.

‍

• Proven scalability — The foundation allows for the addition of new use cases at a reduced marginal cost, and to push the Alteryx product to its limits in partnership with Prime Analytics.

‍