AI-Augmented Testing: The Future of Software Testing
Picture the scene. It’s the 1990s, and it’s the first day of my summer vacation job at a large telecommunications company. My manager hands me a huge book of test requirements and associated test cases and tells me that my next six weeks will be spent using them to test a PABX (Private Automated Branch Exchange) system and some VoIP (Voice Over IP) phones.
So that’s what I did. I spent six weeks sitting in a lab for 12 hours a day, running thousands of manual tests. I wasn’t a domain expert, so I wasn’t sure what PABX or VoIP meant. I wasn’t even an expert manual tester with a stack of testing patterns. I was just a rookie following each test step, expected result and writing out the actual test result.
At the end of the six weeks, I went back to my manager. I told him I’d completed all the tests, and none had failed. I assumed he’d be happy his software had passed every single test. As you can guess what he said. “I know. We’ve run them every year for the last decade. The tests have never failed.”
But because I didn’t know any better, my heart sank. At the time, it felt like I’d just wasted my entire summer. I hadn’t added any value because I’d only proved something that had always been true was still true. None of my work had revealed anything new. All I’d done was take a horrible chore off someone else.
Automation was the Future Once
Once my disappointment had faded, I determined that I would automate the process, so it never had to be done manually again. I spent the next three years fully automating the tests so they could be run more frequently and more quickly. Most importantly, they would never have to be done manually again.
Until very recently, automated testing has been our best possible solution to maximizing test coverage, freeing testers to focus on work that adds more value. What’s changed in the past year or so, of course, is the advent of Generative AI and Large Language Models (LLMs).
Generative AI is the Future Now – But Not as We Know It
I’m guessing that, like most people, you’ve spent the past year or so exploring what gen AI has to offer. You might be using it to write code. You might be using it to write essays or articles. You might be using it like you used to use Google to get answers to questions.
It’s an extraordinary tool that clearly has huge potential for testers – but in its current form it has huge limitations.
It has no domain-specific knowledge and doesn’t know any of the answers. For now, it won’t tell you where it’s got information from so, without your own domain-specific knowledge, you have no way to assess the validity (or otherwise) of the response. This will undoubtedly change in the next year as the EU AI Act comes into force, forcing LLMs to reveal their sources and give data owners the power to restrict the use of their data.
It doesn’t offer any data privacy either. Doubtless we’ve all seen the stories of software engineers feeding proprietary code into ChatGPT and then finding the code has been used as training data for the system.
The reality is that general tools like ChatGPT don’t have much (if any) practical application in our world right now. It doesn’t know what you know so it’s in effect a generalist advising a specialist. You can’t use it as a co-pilot because you can’t – or shouldn’t – feed it any proprietary data or mission-critical information.
Which all leaves us with one question. If gen AI is the future we’ve been told it is, how will it work? The answer is Keysight’s Eggplant Generative AI (GAI).
AI-Augmented Testing - A Testing Expert and an Industry Expert in One
Eggplant Generative AI (GAI) enables testers to take advantage of next generation AI-Augmented Testing capabilities that can actually be used today to make a real difference in Testing & QA.
It’s a highly tuned large language model that’s a true testing expert. It’s trained on ISO/IEEE/BSI/ISTQB® testing material from trusted sources, so you can be confident of the quality of the inputs.
It’s also trained on industry verticals such as healthcare, telecommunications, or aerospace and defense to offer industry-specific insight.
Not only that, but it will also give you its sources too, so you can apply your own knowledge to confirm the quality of the response.
You can use this base model to fine-tune it to your precise requirements. And you can do it with complete confidence because GAI is offline and therefore 100% secure.
Use Your Proprietary Knowledge with No Risk
The offline nature of GAI is super-important because it means you can feed it your IP-sensitive data. You can give it your confluences, specification documents, and entire test repository, safe in the knowledge it will all stay within your own environment.
It gives you a tool that understands testing, your industry vertical, and your specific environment. It can generate fully automated tests and optimize and prioritize test execution through a digital twin model-based testing approach.
It’s a powerful proposition.
The Power of GAI
Think about the domain expertise of testers, manual testers, and exploratory testers with the power of the GAI and AI-Augmented Testing behind them.
GAI can generate all the automation assets, all the models, and all the tests that are needed.
It can go further too, with Sentient Test Expert (STE) from Keysight’s Eggplant.
The Power of STE
You can give STE any digital interface, and it will utilize next-generation cognitive reckoning technology to understand what needs to be tested and why. Human testers can assess the test suggestions, apply their own knowledge to discard the ones that add no value, and ask STE to run the ones that have possibilities.
GAI and STE from Keysight’s Eggplant is What the Future of Testing Really Looks Like – And They’re Here Now
GAI and STE won’t and can’t replace human testers for the same reasons that automation won’t and can’t replace human testers.
But think back to the teenage me who conducted manual tests in a lab for six weeks. Or too far too many manual testers who are still tasked with similar things today. GAI and STE supercharge and give superpowers to these manual testers. They help these manual testers provide – and get – far more value from their work.
GAI and STE launch at Automation Guild 2024. Reserve your spot here. Or, to talk to one of our experts about what GAI and STE could bring to your organization, contact Eggplant today.