Return on investment (ROI) compares an investment’s net benefit with its cost, usually as a percentage: (benefit − cost) / cost × 100. In financial AI, it connects model results to budget and deployment decisions. section 2.4 section 14.6.3
AI maturity shapes the scale of returns:
- Explorer pilots can produce limited returns while testing an initial use case. section 2.4.1
- Integrating AI into one core workflow can increase returns. section 2.4.1
- Embedding AI across enterprise processes is associated with the highest returns. section 2.4.1
Financial value in practice
A GenAI assistant shortened client-proposal turnaround by 20–25%, supporting client growth. An AML system reduced false positives by 60%, while loan-decision automation reached up to 80%. These outcomes connect ROI to revenue, analyst time, and operational workload. section 2.4.2
Listing 14.20 ROI calculation
def calculate_roi(experiment_results, total_high_risk_users=6796,
avg_user_aum=10000, annual_revenue_rate=0.005,
monthly_agent_cost=3000):
"""Calculate ROI for the Retention Agent"""
monthly_control = experiment_results['control_churn_rate']
monthly_treatment = experiment_results['treatment_churn_rate']
annual_control = 1 - (1 - monthly_control) ** 12
annual_treatment = 1 - (1 - monthly_treatment) ** 12
users_saved = total_high_risk_users * (annual_control - annual_treatment)
aum_preserved = users_saved * avg_user_aum
revenue_preserved = aum_preserved * annual_revenue_rate
annual_cost = monthly_agent_cost * 12
net_benefit = revenue_preserved - annual_cost
roi_pct = (net_benefit / annual_cost) * 100
return {
"users_saved_annually": round(users_saved),
"revenue_preserved": round(revenue_preserved),
"annual_cost": annual_cost,
"net_benefit": round(net_benefit),
"roi_percentage": round(roi_pct, 1)
}
roi = calculate_roi(results)
print(f"Users saved: {roi['users_saved_annually']:,}")
print(f"Revenue preserved: ${roi['revenue_preserved']:,}")
print(f"Annual cost: ${roi['annual_cost']:,}")
print(f"ROI: {roi['roi_percentage']}%")
This calculation annualizes monthly churn rates, estimates users and assets under management preserved, converts those assets to revenue, and compares revenue with annual agent cost. section 14.6.3
