ANZ Bank Github CoPilot Breakthrough
Discover how ANZ Bank’s use of GitHub Copilot led to a breakthrough in productivity and innovation in the banking sector.
• ANZ Bank’s integration of GitHub Copilot has increased software engineering productivity by over 40%.
• Generative AI in banking enhances customer interactions, accelerates code development, and optimises data utilisation.
• Global banks adopt centralised AI models to streamline implementation and maximise impact.
In the rapidly evolving banking landscape, generative AI is a transformative force. ANZ Bank’s innovative use of GitHub Copilot, a generative AI tool, has significantly enhanced productivity and code quality among its engineers. This development not only underscores the potential of AI in banking but also sets a benchmark for other financial institutions worldwide.
The ANZ Experience
ANZ Bank conducted a six-week experiment integrating GitHub Copilot into their software development processes. The results were impressive, showing a 42.36% average increase in productivity. As stated in their research,
“The group that had access to GitHub Copilot was able to complete their tasks 42.36% faster than the control group participants”.
This experiment involved over 100 engineers working on Python coding challenges. Copilot users consistently outperformed those without the AI tool.
Beyond productivity, Copilot also improved code quality, reducing bugs and code smells. Engineers reported a positive sentiment towards using Copilot, appreciating its ability to generate relevant code snippets and assist with debugging and documentation.
Global Perspectives
While ANZ’s approach focuses on enhancing software engineering, other banks worldwide also leverage AI in various ways. A study by McKinsey highlights that banks using centralised AI models are leading the charge in implementing generative AI. About 70% of these institutions have advanced beyond pilot stages, integrating AI use cases into production. This centralised model facilitates faster skill and capability development, ensuring a cohesive strategy across the organisation.
For instance, European and US financial institutions have adopted centralised AI frameworks to streamline operations and manage risks effectively. These frameworks allow banks to prioritise AI use cases that align with their strategic goals, ensuring efficient resource allocation and value creation.
Comparative Analysis
The paper “Refactoring vs Refuctoring: Advancing the State of AI-Automated Code Improvements” by Adam Tornhill and colleagues, provides an insightful contrast to ANZ’s findings. This study benchmarks popular Large Language Models (LLMs) for their ability to refactor code, revealing significant challenges. They found that existing AI solutions deliver functionally correct refactorings in only 37% of cases. However, introducing a novel fact-checking innovation improved accuracy to 98%, significantly enhancing code quality and mitigating technical debt.
Similarly, the study “GitHub Copilot AI Pair Programmer: Asset or Liability?” by Arghavan Moradi Dakhel and colleagues explores Copilot’s performance in solving fundamental algorithmic problems and comparing its solutions to those of human programmers. The study found that while Copilot can generate correct solutions for most problems, it often produces buggy or non-reproducible code. The authors conclude that Copilot can be an asset when used by experienced developers but may become a liability in the hands of novices who may not adequately filter its suggestions.
Key Takeaways for CIOs
1. Strategic Implementation
• Centralised Model: Adopt a centralised AI model to streamline AI initiatives, ensuring a cohesive strategy and efficient resource allocation.
• Skill Development: Invest in training programs to equip your team with the necessary skills to leverage AI tools effectively.
2. Risk Management
• Fact-Checking Mechanisms: Implement robust fact-checking mechanisms to enhance the accuracy and reliability of AI-generated solutions.
• Experienced Oversight: Ensure experienced developers oversee the use of AI tools to mitigate risks associated with buggy or non-optimal code.
3. Customer Experience
• Personalisation: Use AI to provide personalised and timely services, enhancing customer satisfaction and loyalty.
• Data Utilisation: Leverage AI to optimise data utilisation, offering customers valuable insights and forecasts.
Call to Action
For CEOs and business leaders, integrating generative AI into your operations is not just an option but a necessity. The benefits — from enhanced productivity and code quality to improved customer interactions and robust risk management—are too significant to ignore.
Here are actionable steps to consider:
1. Invest in AI Training: Equip your teams with the necessary skills to leverage AI tools effectively.
2. Adopt a Centralised AI Model: Streamline AI initiatives across your organisation to maximise impact and ensure cohesive implementation.
3. Focus on Customer Experience: Use AI to provide personalised and timely services, enhancing customer satisfaction and loyalty.
Embrace the potential of generative AI and lead your organisation into a future where innovation drives growth and success. The time to act is now.
