AI Usage Frameworks and how bank are using this to Transform Banking Operations
As AI continues to reshape the financial services landscape, understanding how to strategically adopt it can give banks a competitive edge. Today, I’m diving into six AI usage frameworks that are helping banks worldwide enhance efficiency, improve customer experiences, and manage risks.
1. Microsoft AI Maturity Model
What It Is: This framework tracks AI adoption across three stages: Assisted Intelligence (AI supports humans), Augmented Intelligence (AI enhances human decisions), and Autonomous Intelligence (AI operates independently).
HSBC has leveraged this model to evolve its AI capabilities. Initially, HSBC used Assisted Intelligence to analyze transaction data and flag potential fraud for human review. As they progressed to Augmented Intelligence, they introduced AI tools to enhance customer service, such as chatbots that assist agents with real-time insights during client interactions. Now, HSBC is moving toward Autonomous Intelligence with AI-driven anti-money laundering systems that independently detect and report suspicious activities, reducing manual oversight while ensuring compliance.
Takeaway - Start with AI as a support tool and gradually scale to autonomous systems, ensuring robust governance at each stage.
2. PwC AI Augmentation Spectrum
What It Is: This framework categorizes AI roles by complexity: Executor (performs tasks), Self-Learner (improves over time), and Decision-Maker (makes autonomous decisions).
JPMorgan Chase has applied this spectrum effectively. As an Executor, their COiN platform uses AI to process legal documents, extracting key data points from contracts in seconds—a task that previously took hours. As a Self-Learner, JPMorgan’s AI models for credit risk assessment continuously improve by learning from new customer data, refining predictions over time. As a Decision-Maker, their AI systems now autonomously execute trades in high-frequency trading environments, making split-second decisions based on market trends.
Use AI to automate routine tasks first, then let it learn and eventually take on decision-making roles in controlled environments.
3. Deloitte Augmented Intelligence Framework
What It Is: This framework focuses on three pillars: Automate (handle repetitive tasks), Augment (enhance human capabilities), and Amplify (speed up decision-making).
Real-World Example: Deutsche Bank has adopted this framework to streamline operations. They Automate routine processes like transaction reconciliation, where AI processes thousands of payments daily, reducing errors. For Augmentation, Deutsche Bank uses AI to assist relationship managers by providing real-time customer insights, helping them offer tailored investment advice. To Amplify, their AI systems analyze market data to accelerate risk assessment for loan approvals, enabling faster decisions without compromising accuracy.
Leverage AI to free up human resources from repetitive tasks, enhance advisory roles, and speed up critical decisions.
4. Gartner Autonomous Systems Framework
What It Is: This framework maps AI autonomy against human involvement: None/Low (AI-led), Semi-Autonomous (shared tasks), and Fully Autonomous (human-driven with AI support).
Bank of America has utilized this framework to enhance its operations. With None/Low human involvement, their AI-powered virtual assistant, Erica, autonomously handles customer queries like balance checks and bill payments, serving millions of users. In a Semi-Autonomous setup, Bank of America’s AI collaborates with human analysts to predict market trends, combining AI insights with expert judgment for investment strategies. For Fully Autonomous tasks, their AI systems assist compliance teams by automatically generating regulatory reports, with humans overseeing the final submission.
Balance AI autonomy with human oversight based on task sensitivity, ensuring compliance in highly regulated areas.
5. MIT Institute of Technology Human-in-the-Loop (HITL) AI Mode
What It Is: This model emphasizes a feedback loop where humans provide input, AI processes it, and humans review the output to improve accuracy.
Standard Chartered Bank has implemented HITL to enhance fraud detection. Humans first define patterns of suspicious behavior, which AI uses to scan transactions. The AI flags potential fraud cases, and human experts review these outputs to confirm accuracy. Feedback from these reviews is fed back into the AI, improving its detection algorithms over time. This iterative process has significantly reduced false positives in their fraud monitoring systems.
Use HITL for high-stakes tasks like fraud detection, where human expertise can refine AI accuracy over time.
6. Harvard Business Review Human-AI Teaming Model
What It Is: This model defines AI as a Tool (provides insights), a Collaborator (shares tasks), or a Manager (leads processes).
Citibank has embraced this model across its operations. As a Tool, Citibank’s AI analyzes customer spending patterns to provide insights for personalized product offerings, like suggesting a new credit card. As a Collaborator, their AI works with human traders to optimize algorithmic trading strategies, combining AI’s speed with traders’ market knowledge. As a Manager, Citibank uses AI to oversee portfolio risk management, automatically adjusting investments to minimize losses based on market volatility, with humans setting the initial parameters.
Deploy AI in roles that match its strengths—supporting, collaborating, or managing—depending on the complexity and risk of the task.
These frameworks highlight the versatility of AI in banking, from automating routine tasks to driving strategic decisions. The key is to adopt a phased approach, starting with support roles and scaling to autonomous systems, while maintaining strong governance to address regulatory and ethical concerns.
#AIinBanking #FinancialServices #Innovation #DigitalTransformation #BankingTech
Deepanjan