AI can help banks unleash a new era of software engineering productivity

How AI could boost engineering productivity

Multiple studies have noted that large language models are helping programmers code faster: Productivity increases range from 30% to 55%.⁷ Within the banking industry, Citizens Bank observed a productivity boost of 20% among a group of engineers using generative AI in a test case.⁸

In our view, this increase in productivity should be demonstrated across software delivery. AI is transforming the speed and quality of code generation and playing an important role in requirement analysis by automatically identifying and classifying requirements, detecting tacit requests, and building a comprehensive problem statement. They are helping make software design and testing processes more efficient. System engineers can use natural language processing models to feed requirements into AI systems, which can then propose initial design options and their respective trade-offs. Some engineering teams are using AI testing tools, such as Applitools, Functionize, and Testim, to help automate test case generation and execute a large number of test cases in a short period of time.⁹ These programs can also learn from past tests to improve test quality.

And then there are AI assistants that can now deploy and maintain software. For instance, AI algorithms are automating continuous integration and deployment to help optimize deployment schedules, reduce downtimes, and confirm rollouts of new features and updates.

These comprehensive improvements should result in a quicker understanding of client or business requirements, heightened design quality, superior testing, predictive maintenance cycles, and enhanced system stability. Ultimately, they should lead to better software engineering performance.

The challenges in software engineering at banks

According to Gartner®, enterprise IT spending on software by the banking and investment services in the United States reached approximately US$107.8 billion in 2024.¹⁰ Overall, US banks employ around 100,000 employees for software development.¹¹ Some large banks have as many as 15% to 25% of their total workforce involved in software-related tasks.¹²

Despite these investments, the current state of software engineering in the banking industry likely has enormous room for improvement. According to Deloitte interviews of different banks’ technology or software engineering heads conducted in 2024, banks’ engineering projects are often inefficient¹³ and weren’t built for scale, which can result in costly maintenance, integration issues, and slower runtimes.¹⁴ (See our report “How should banks respond to the current disruption in software engineering?” for more details.)

AI tools can help address these and other challenges by improving engineering productivity at banks in multiple ways. Take, for instance, writing or maintaining code for mainframes. Some generative AI models are being trained to rewrite the 1960s-era code that underpins older cores to be compatible with modern software. IBM’s watsonx Code Assistant can read, understand, and rescript COBOL-coded applications in Java.¹⁵ Citigroup’s 30,000 developers are incorporating generative AI coding assistants in their modernization efforts.¹⁶ Furthermore, the low code and no code functionalities of many AI systems are already helping accelerate development cycles.¹⁷

As these experiments mature, banks should be able to see the business impact more clearly. In fact, these AI models will only get more advanced and valuable with time, opening up new possibilities to write faster and better programs and yield even higher productivity gains.¹⁸ For example, AI agents can now autonomously convert ideas expressed by humans in natural language into executable code.¹⁹ Models launched recently, which are often less expensive and use fewer resources, could unleash more cost-efficient AI tools for software engineering.²⁰

Strategies to help maximize productivity gains

Here are some actions software engineering leaders can consider taking as they start scaling AI across the software development life cycle.

1. Identify where inefficiencies are the highest and persistent, and then deploy AI tools in those areas first.
2. Communicate directly with software engineers and developers about potential job shifts, upskilling opportunities, and new roles emerging from AI adoption. Providing regular updates, open forums, and clear career development pathways can also help alleviate employee concerns and foster a positive attitude toward AI.
3. Develop robust governance models to oversee AI integration. Executives should establish a framework that includes clear policies and addresses potential risks such as data privacy breaches, algorithmic bias, and operational disruptions.  
4. Factor in third-party vendors as part of the implementation strategy. This can be especially critical for banks that lack the scale and infrastructure to build in-house AI tools.
5. Share plans and progress with stakeholders within and outside the organization on how AI’s capabilities are being incorporated into software development, and the governance measures that are being put in place to manage risks.
6. Build connected teams across business areas to help oversee AI integration processes and foster collaboration among business heads. 

This article contains general information and predictions only and Deloitte is not, by means of this article, rendering accounting, business, financial, investment, legal, tax, or other professional advice or services. This article is not a substitute for such professional advice or services, nor should it be used as a basis for any decision or action that may affect your business. Before making any decision or taking any action that may affect your business, you should consult a qualified professional advisor.

Deloitte shall not be responsible for any loss sustained by any person who relies on this article.