Skip to content

Table of Contents

    Addressing Tech Leaders' Concerns About AI's Role in the Modern SDLC

    White Blue Black Minimalist Artificial Intelligence YouTube Thumbnail (4)

    On 25th July, our virtual event, “Expert Roundtable: AI's Role in the Modern SDLC“, garnered immense interest and enthusiasm from C-level executives across the Asia-Pacific region. This event was a testament to the escalating significance of AI in the Software Development Life Cycle (SDLC).

    From the invaluable insights shared by our esteemed panel of experts - David McKeague, Paul Steven Conyngham, and Du Hà, attendees can understand more about AI use cases, ideas for implementing AI in software development, the strategic roadmap for successful AI integration, and prompt engineering in the SDLC.

    During the webinar, there were some concerns voiced by tech leaders regarding AI’s application in the modern SDLC. By examining real-world use cases and best practices, here’re insightful answers needed to embrace AI as a powerful ally in the business software development journey.

    1. How to implement AI in the SDLC? Which stage of the software development process should apply AI?

    Implementing AI effectively in the SDLC requires a thoughtful approach and a well-defined strategy. Here are some critical steps to consider:

    • Determine which areas of the SDLC could benefit the most from AI.  It’s worth considering concentrating on use cases that align with your organisation's objectives, such as automating repetitive tasks, improving code quality, or enhancing testing processes.
    • Ensure your business has reliable and sufficient data to train the AI models effectively. Data preparation and data governance are key aspects to address before AI application.
    • Collaborate closely with the dedicated software development team to ensure they understand the AI implementation strategy and have sufficient skills to foster a smooth transition.

    Throughout seven critical stages of the SDLC process, AI can be applied to ensure the productivity and delivery of high-quality software. Here are some examples of implementing Ai in different phases of SDLC:

    • Requirements gathering: AI can assist in analysing vast amounts of data, user feedback, and market trends to generate comprehensive and accurate software requirements.
    • Planning documentation: AI algorithms can evaluate the time needed for each development phase, ensuring realistic and achievable project timelines.
    • Software development: A prominent instance of an AI-powered tool is ChatGPT since you can use it to generate codes in different languages and frameworks and suggest best code practices.

    Read the full article: An Ultimate Guide to Applying AI in Software Development Lifecycle Process

    2. Are there any use cases of Chatbots for the BFSI sector? 

    Chatbots are gaining popularity in the BFSI industry to deliver efficient and cost-effective customer service. Mobile app users can now engage with chatbots to gain information about their banking inquiries, as well as resolve concerns related to accounts, transactions, or products.

    By offering round-the-clock customer support, Chatbots can help banks and financial institutions significantly reduce wait times and improve customer satisfaction. In addition, chatbots can handle multiple queries simultaneously, allowing them to efficiently handle a large volume of customer requests. For example, Bank of America's virtual assistant, Erica, serves as a versatile chatbot assisting customers with tasks such as checking account balances, making transfers, and even disputing charges.

    Besides chatbots, there are also other implementations of Generative AI in the BFSI sector, including synthetic data generation, fraud detection, financial forecasting, etc. 

    3. How to implement AI in existing Investment Platforms and strategies to get started with this?

    Regarding investment software use cases, AI algorithms can be utilised to offer personalised recommendations based on the investor’s risk tolerance, from that optimising their investment strategies. Applying AI to investment platforms requires a systematic approach and careful consideration of various factors. To implement AI effectively in investment platforms, here are key steps that you should consider:

    • Defining objectives: It’s essential to identify the specific challenges you aim to address (i.e. optimizing portfolio allocations, improving risk management, or offering personalised investment recommendations) at the beginning to tailor the AI integration that is best suited for your platform's unique needs. 
    • Data gathering and preparation: Data is the lifeblood of AI-driven solutions. Gather and clean relevant data from diverse sources, including market data, economic indicators, company financials, and investor behaviour, so that AI algorithms can identify patterns, correlations, and trends, enabling sophisticated analyses and predictions.
    • Choosing AI techniques and developing AI models: Choose the appropriate AI techniques that align with your investment objectives and data and build robust AI models using the selected techniques.
    • Integrating AI with Investment Platforms: To ensure AI's seamless integration, invest in the proper infrastructure and technology stack. Facilitate communication between AI factors and other systems to enable real-time decision-making and streamline processes.

    4. How can AI help large organisations such as banks manage and reduce the complicated growth of software?

    Currently, the majority of banks are staying at the legacy system and integrating AI will help it improve customer experience and attract more younger users. With the upsurge of mobile app users, it’s essential to handle and centralise such a vast amount of data effectively while minimising the risk of data loss. 

    AI-powered algorithms can gather and analyse large data quickly and identify trends and correlations that may not be apparent through manual analysis. This helps banks gain valuable insights into customer behaviour, market trends, and risk factors. With a data-driven practice, banks can make well-informed decisions to manage and reduce software complexity effectively.

    Besides, as an extended use case of AI, Autonomous Testing is the next generation of software testing approaches that can assist banks in performing parallel testing and enabling multiple tests to run simultaneously on different environments. 

    We also notice that large organizations, including banks, are deeply concerned about cutting operational expenses and mitigating risks. To address this, they are turning to Robotic Process Automation (RPA), which utilizes software to emulate rule-based digital tasks typically performed by humans. In the banking sector, RPA is employed to streamline processes and eliminate time-consuming and error-prone tasks related to entering customer data from various sources like contracts and forms.

    5. What are the different tools or procedures that can be used for different AI tasks, like automating tests for business or code review or generating requirements?

    There are numerous diverse tools that cater to specific scenarios that should be considered. Generally, businesses can use ChatGPT and GitHub Copilot. The key is that nobody should completely rely on the outcomes of these AI tools. We need to justify the outcomes carefully, as we're the ones creating them. Here are some use cases that you can consider.

    • ChatGPT can help write user stories in SDLC, and you might find out that it helps provide more thoughtful scenarios than you can expect.
    • ChatGPT can help generate automated test scripts in Selenium and much more
    • GitHub Copilot can suggest ways to improve the code, such as making it more efficient, idiomatic, or readable.
    • GitHub Copilot can identify potential bugs in the code, such as typos, logical errors, or security vulnerabilities.
    • GitHub Copilot can provide feedback on the code, such as whether it is well-organized, well-documented, or easy to understand.

    6. Given GPT tools don’t actually have an understanding of the problem they give solutions to, can this end up with false confidence in generated code, resulting in emergent errors?

    This issue, also known as AI hallucination, is extremely dangerous for business. Without validation, we shouldn't put our faith in AI-generated results.  Additionally, this holds true for all other SDLC processes as well, not only those involving AI technologies. 

    This can lead to false confidence in the generated code, which may look plausible but could contain subtle errors or vulnerabilities that are not immediately apparent. These errors may only emerge when the code is executed or integrated into a larger system, potentially causing unexpected behaviour or security risks.

    As an illustration, a developer might have great confidence in his code, but the testing process is always in place to help him validate and present alternative viewpoints. Every workflow should be designed with the standard process in mind. And this in no way implies that we lack faith in individuals. It's how business is conducted. 

    7. How will businesses that still have many legacy tools be able to maximise the speed of AI when their anchors in legacy are likely to determine the speed at which they can run?

    Legacy systems designed with a human interface in mind may slow down the speed of AI. This is because legacy systems are often not designed to be scalable or efficient, and they may not be able to handle the large amounts of data that AI requires. 

    One of the biggest bottlenecks for AI is dirty data. Legacy systems often have a lot of dirty data, which can slow down AI models. Businesses can improve the speed of AI by cleaning up their data.

    Businesses can use modern technology to construct new, standalone features that have minimal impact on the existing system rather than attempting to patch or replace the entire legacy one. For instance, the bank will often need its users to present themselves in person at the bank office for verification in order to register a user account. A new component, such as eKYC (Electronic Know Your Customer), which enables the bank to validate its users remotely, can be built or integrated. 

    8. Is there any real-world example of AI in the BFSI industry?

    Several banking institutions have already leveraged AI technologies to enhance their services, detect fraud, and improve customer experiences. Here are a few real-world examples from banks and financial institutions in Australia:

    • AI has long been the foundation of CommBank app features like Bill Sense, enabling the prediction of bills and cash flow forecasts and  NameCheck, scam detection initiatives underpinned by AI.
    • Brighte integrated Curious Thing’s conversational AI solution - a service provided by our speaker David McKeague - to increase answer ability and better manage security risks.

    In addition, banks in the ASEAN region are also in a race to implement AI in their legacy systems:

    • DBS Bank is one of the frontiers in implementing RPA to automate repetitive and time-consuming tasks such as customer onboarding and credit analysis.
    • By integrating with AI Voice Banking technology, the TPBank VoicePay feature allows customers to perform many touchless transactions efficiently while allowing the bank to recognize and understand the intention of users, providing a seamless experience for users.

    The aforementioned questions are among the frequently asked ones we received during and after the webinar. If you are interested in delving deeper into this cutting-edge technology and implementing it into your current business system, our team of experts is readily available to offer assistance!

    Contact Our Experts for AI Initiatives