How leveraging the newest and most powerful AI tools available can accelerate productivity to levels never previously imagined.

“To err is human, but to really foul up requires a computer!”

This old saying is just as true today as it was when it was first invented. However, with a proper testing strategy and approach, we can significantly reduce or even eliminate the risk of it becoming a reality. With the surge of AI-based tools, due to the increased availability of cheaper/faster computational resources, we have seen their benefits and just how transformational they can be.

There are millions of web results available for a simple search for information on AI, and many very thick books have been written by PhD-holding academics. Within AI testing, there are many new concepts (from the definition of binary classifiers to neurons and neural networks and more) along with a plethora of new and often confusing terminology (e.g., 5 different definitions of ‘bias’ depending on context) to learn. Each of these require a full 2-day course just to gain a foundational knowledge (for example, ISTQB AI Testing).

To learn more about how banks can manage the best AI outcomes for cost, sustainability and business value, check out this blog.

Rather than being overwhelmed by all this detail about AI, its design, and implementation, we should be most interested in specific use cases where AI tools will add significant value and ROI to testing and quality within the IT project systems development life cycle (SDLC).

This area is of specific importance to financial IT projects, which are often planned with tight deadlines due to business needs and with short development phases (especially so in Agile projects with two-week sprints). The ‘how’ and ‘where’ of AI’s application in testing will be the most important change in QA strategy since test automation became a buzzword.

Use cases

Here are three specific use cases that most people with IT SDLC experience can relate to:

1. Requirements Analysis 

Large language model (LLM)-based tools such as ChatGPT can produce testable use cases from a given set of requirements, due to their extensive NLP (natural language processing) capability, which has developed and grown over time.

Simpler, testable use cases are often missing from requirements, but they are the foundation of good test cases and test conditions. They must be identified and documented to ensure that each requirement has been interpreted and delivered correctly.

It’s also possible that these AI-generated testable use cases will help flag anomalies and gaps that humans struggle to spot when provided only with a compound requirement.

2. Development

The code produced by AI code generators may not always be accurate for simple or repetitive tasks. However, from a testing point of view, AI code generators can deliver far more value by writing unit tests for already-written code.

This can be achieved by analysing the code and generating a suite of tests that fully cover the boundaries of the values to be tested.

This allows developers to deliver much higher-quality products by shifting the more mundane task of writing test cases to an AI tool. Or even, when applied to legacy code, to create unit tests that had previously been omitted due to time/resource constraints.

3. UI Testing

AI-powered record/replay tools are the biggest innovation in this area. In the past, we only had simple tools that could record and replay user interaction, either by cursor position or by identifying the buttons on the screen (e.g., QTP or any script-based framework).

These technologies are prone to creating ‘brittle’ automated tests, which would ‘break’ often and easily if any slight change occurred in the UI or to a button label.

AI-powered UI tools, in contrast, can recognise the images and determine what the user is trying to do. For example, if the ‘Submit’ button was moved or even renamed to ‘OK’, an AI-powered tool will still find the right button and select it during test execution. A feature known as ‘Self-Healing’.

A great example is a browser-based trading platform. Creating a single UI test that contains 20 steps can take 2-3 hours (using Selenium with JavaScript) due to the complexity of finding the correct button labels in each Test Step. Using an AI-powered record/replay tool, the same test can be created in just 10-15 minutes, and it would be significantly more reliable!

In summary

There are many more potential applications for each of the hundreds of AI-powered tools available for improving and accelerating QA and testing activities. These can benefit every project in some way—whether it’s for greenfield development or for a legacy application with a backlog of issues to be analysed and remediated.

How NTT DATA can help

At NTT DATA, we take pride in our understanding of quality assurance and testing, recognising the immense value they bring. Our expertise in banking and the financial markets segment enables us to understand our client’s challenges with new projects and legacy platforms that need maintenance or upgrades.

If you want to learn more about our approach to QA & testing, then please get in touch