InsightCenter – Anticipate and mitigate operational losses

Centerprise's InsightCenter leverages AI with the goal of operational loss mitigation. Mining a variety of data sources to train a collection of AI models and ML algorithms, InsightCenter assesses the risk of future losses for each business unit, and provides insight on the drivers of that risk, so that meaningful mitigation steps can be taken.

Think of these drivers as "hidden KRIs" – specific to each individual business unit – that the AI discovers, and that are not intended to replace traditional KRIs estabished by your Risk team but rather to supplement them. Your Risk team can then weight the scores produced by the AI as they see fit, and take the mitigation actions that they deem appropriate.

Data Sources

• Operational loss and near-miss history from OpRiskCenter or alternate internal sources
• RCSA ratings, residuals, and control test / failure statistics from OpRiskCenter or alternate source
• Issue and action plan data (issue ages, time to resolution, etc.) from OpRiskCenter or alternate source
• Incidents, complaints, IT outages, vendor dependencies, etc. from the appropriate internal systems
• Staffing ratios, average tenure, turnover from internal HR system
• NLP embeddings derived from loss descriptions, RCSA commentary, issues/incidents/tickets, customer service chatbot transcripts, and other unstructured text sources

Algorithms & Models

• Normalization, categorization, time-windowing and trending for feature entineering
• DBSCAN and k-Means clustering for unsupervised learning
• Gradient Boosted Machines (XGBoost), Random Forests, neural nets for structured data
• NLP (TF-IDF, DistilBERT, MiniLM, NER) for unstructured (text) data

Outputs

• Dashboards, heat maps, trend charts of risk scores by business unit, Basel category, and time period
• Risk driver analysis – top N drivers, risk contribution scatter plots
• Training statistics and parity plot for validation / assessment of model training