Case study:

AI Agent Applied in Banking Commercial operations

EXECUTIVE SUMMARY

Converting banking customers to loan products requires marketing teams to spend weeks on segmentation, BI teams to build lists, and sales teams to contact thousands
of prospects. This case study shows how a combined ML/LLM architecture automates prospect targeting and content generation, reducing campaign time from 4 weeks to 14 days while maintaining human-in-the-loop.

1 Detecting propensity to buy

Utilizing advanced analytics on user data to identify clients most likely to be interested in new loan products.

2 Targeted segmentation

Extracting a subset of high-propensity users to create a focused group of sales targets, optimizing resource allocation.

3 Personalized outreach

Tailoring communication and offers based on clients’ previous buying activities and preferences, enhancing relevance and appeal.

4 Empowering sales teams

Providing sales representatives with additional, actionable information to improve the quality and effectiveness of client interactions.

AT A GLANCE
Challenges:
  • Inefficient Conversions
  • High Operational Costs
  • Slow Time-to-Market
  • Suboptimal Targeting
Challenges:
  • 56% Faster
  • 69% More Efficient
  • 10x Smarter Marketing

This AI-driven process significantly improved the bank's marketing efficiency and sales performance, leading to higher conversion rates and reduced operational costs.

Chief Commercial Officer

KEY PERFORMANCE INDICATORS (KPIs):

Time-to-market reduction: 56%

Decreased from 4 weeks to 14 days

Resource efficiency improvement: 68%

Reduced man-day requirements from 16 MD to 5 MD

Marketing targeting efficiency: Contacted 10X less people, delivering same results

Sales agents contacted 5,000 customers instead of 50,000, delivering the same results.

PROBLEM BREAKDOWN

The bank’s traditional approach to loan product marketing and sales was time-consuming, resource-intensive, and yielded suboptimal results. The process typically involved:

  • Commercial teams (Marketing & Sales) receiving a task to sell more of X loans
  • Marketing team spending time on outlining the right persona and building A, B, C segments (2 MD)
  • Internal data analytics team (Business Intelligence) building segments (2 MD)
  • Marketing evaluating and aligning on the outputs (1 MD)
  • Drafting personalized emails and sales scripts (3 MD)
  • Sending out emails, briefing sales team, and making calls (10 MD)
This process took 3-4 weeks and 16 man-days to execute, resulting in:
  • High operational costs
  • Slow time-to-market for campaigns
  • Suboptimal targeting and personalization
  • Inefficient use of sales team resources
  • High percentage of false positives

AI Agent Applied in Banking Commercial Operations

This case shows a simplified AI agentic arhitecture where ML ranks prospects by propensity to buy, LLM generates personalized outreach, and humans review content quality - while maintaining human-in-the-loop.

Solution: AI Agents for Commercial Back Office Work

We implemented AI agents to streamline and optimize the back-office work. The system comprises two main components: Machine Learning and Large Language Model (LLM)

1. Machine Learning AI

This component analyzes historical data to prepare propensity to buy model and build targeted lists. Model calculates the probability to buy loan for all customers that comply with business rules.
Performance boost:

  • Eliminated the process of marketing team brainstorming segments
  • Removed dependency on internal BI team for segmentation
  • Reduced list creation time from 5 MD (2 MD for persona outlining + 2 MD for BI team + 1 MD for evaluation) to near-instantaneous processing
  • Higher success rate because campaign only focus on customers with high probability
How: ML Model Deployment and Operation

Data Processing:

  • Monthly data extraction from the client database to Azure ML Studio
  • Feature calculation using the last 12 months of data

Prediction and Ranking:

  • XGBoost algorithm predicts conversion probabilities for all products
  • Candidates ranked for each product based on these probabilities

Campaign List Optimization:

  • Match of prediction lists with campaign candidates
  • Computation of cumulative attributes for targeted marketing

Continuous Improvement:

  • Periodic model retraining to capture new market dynamics
  • Robust evaluation process to refine model accuracy

Automation:

  • Scripts prepared for process automation
  • Plans for automated explainability and evaluation of predictions

This ML-driven approach enables the bank to make data-informed decisions, resulting in higher conversion rates and improved marketing efficiency.

2. Large Language Model (LLM) AI

Once the list is built, this component uses advanced language models to draft personalized emails and scripts for the sales team.
Performance boost:

  • Marketing team now only needs to review and edit scripts if necessary
  • Resulted in faster design and campaign deployment
  • Reduced content creation time from 3 MD to a fraction of that time (exact reduction not specified in the original data)
Result and Impact

The implementation of AI agents significantly enhanced the overall performance of marketing campaigns:

  1. Process Optimization:
    1. Eliminated unnecessary steps in the processes
    2. Automated list building and email drafting
    3. Reduced campaign execution time from 16 MD to 5 MD (68% improvement)
    4. Shortened overall campaign timeline from 4 weeks to 14 days (56% reduction)
  2. Improved Targeting and Personalization:
    1. Increased accuracy of propensity modeling (exact percentage not provided in original data)
    2. Enhanced personalization based on previous buying activities
  3. Sales Team Empowerment:
    1. Provided sales representatives with AI-generated insights and additional information
  4. Financial Impact:
    1. Reduced marketing campaign costs (exact percentage not provided in original data)
    2. Increased conversion rates (exact percentage not provided in original data)
  5. Overall Efficiency:
    1. Personalized campaigns are now executed in 5 days instead of 16 days
    2. Total process time reduced from 3-4 weeks to 14 days
Conclusion

By deploying an integrated AI solution combining machine learning and large language models, the bank has significantly transformed its marketing and sales processes. This implementation has led to faster campaign execution, reduced man-days, and the potential for higher conversion rates, all while enhancing personalization of customer outreach.

This case study demonstrates the transformative power of AI in traditional banking operations, offering a competitive edge in an increasingly digital financial landscape. As the sector evolves, institutions that effectively leverage AI will be well-positioned to lead in customer satisfaction, operational efficiency, and market responsiveness.

Related Projects