Citi Bank · Q2 2026
AiRi Composite Score
Piloting
- Infra3.80
- Wealth3.71
- MLOps3.55
- Ethics1.67
- Data1.85
- Talent2.17
Domain maturity radar
hover for insightGap-to-target (lowest first)
What it means
ROI was estimated at project kickoff but nobody tracked whether it materialized. Leadership hears 'we saved 2,000 hours' but nobody can tell the CFO whether that translated into headcount reduction, throughput increase, or margin improvement. Hours saved is not value captured. The data science team is stretched across multiple pilots, delivering none at production quality. Data infrastructure supports the pilot use case but breaks when a second domain is added. There is a growing 'pilot zoo' that consumes budget without scaling.
Next steps
Implement standard value measurement with banking-specific baselines: false positive rate before/after, approval rates before/after, cost-per-transaction, time-to-decision, and financial attribution for each. Invest in MLOps foundations: standardized pipelines, model registry, and API-first integration. Consolidate the portfolio — kill low-impact pilots and redirect to 2-3 high-value initiatives with clear production paths. A bank with 3 models in production is more mature than a bank with 15 pilots.
Domain breakdown
| Domain | Maturity | vs. composite | Coverage | Weight |
|---|---|---|---|---|
| AI Strategy & Governance | 3.21 | +0.37 | 8/8 | 9.0 |
| Data & Foundation Architecture | 1.85 | -0.98 | 4/4 | 10.0 |
| AI Talent & Capability Building | 2.17 | -0.67 | 8/8 | 6.0 |
| Model Lifecycle & MLOps | 3.55 | +0.72 | 4/4 | 9.0 |
| Responsible AI & Ethics | 1.67 | -1.16 | 4/4 | 9.0 |
| AI Value Measurement & ROI | 2.75 | -0.09 | 8/8 | 10.0 |
| Change Management & Adoption | 2.77 | -0.06 | 8/8 | 6.0 |
| Enterprise AI Infrastructure | 3.80 | +0.97 | 4/4 | 9.0 |
| Fraud & Financial CrimeSpecialty | 3.26 | +0.43 | 8/8 | 10.0 |
| Wealth & Advisory AISpecialty | 3.71 | +0.88 | 8/8 | 3.0 |
This score is not a generic benchmark. Every domain weight, every rubric threshold, and every scoring band is tuned specifically to how Tier-1 Global/National Universal Banks organisations compete and operate in Banking.
Nearest peer lens: Your AI posture most closely resembles a Super-Regional Banks profile (Δ0.01 when domain weights are re-calibrated for that segment). This is a signal — not a prescription.
Sutherland recommendations
Solid strategic foundation. Board engaged. Focus on optimization and scaling.
Approach: Establish AI CoE with dedicated leadership and capability building
Platform emerging. Gaps in lineage, quality, real-time access. Limited governance.
Approach: Design enterprise data platform with automated quality controls
Team building but attrition risk high. Career paths not defined.
Approach: Launch AI fundamentals training, formalize career progression
Fully industrialized operations. Zero-touch model lifecycle. Predictive maintenance.
Approach: Implement autonomous ML platform with self-healing and auto-remediation
Ethics framework exists but application inconsistent. Manual oversight required.
Approach: Formalize responsible AI governance and implement automated bias detection
Systematic value measurement. Focus on cost optimization and portfolio management.
Approach: Implement AI FinOps for cost per inference and portfolio optimization
Active adoption management. AI embedded in performance evaluations.
Approach: Scale adoption through champion networks and cultural embedding
Enterprise-grade autonomous platform. Self-managing with predictive scaling.
Approach: Implement autonomous infrastructure with AI-driven resource optimization