How Pick n Pay plans to transform software test cases with AI
Blog: OpenText
Headquartered in Cape Town, South Africa, Pick n Pay is a leading retailer that operates more than 1,650 stores in seven countries on the African continent. While we continue to grow our physical footprint, online and mobile shopping are increasingly popular channels for our customers. Alongside our standard home delivery services, we recently launched the ASAP! mobile app for deliveries within 60 minutes of more than 400 participating stores.
The constantly growing demand for new digital tools and new features for existing tools—both from internal business users and from our customers—puts our software testing and development workflows under a lot of pressure. To help us stay on top of the workload, we’re heavy users of OpenText™ solutions, including the latest ValueEdge™ release for automating and managing our software cycle from end to end. We use OpenText MF Connect to sync releases, stories, and bugs between ValueEdge and Jira in real time.
Beta testing DevOps Aviator
After I presented at OpenText World 2023 in Las Vegas, I accepted an invitation from OpenText to take part in the beta testing of its new OpenText™ DevOps Aviator™ solution. Aviator is Large Language Model (LLM) technology that uses Google’s PaLM 2 data lake as its main source of reference data. We use ValueEdge as a SaaS solution, so OpenText simply activated DevOps Aviator in our workspace--that’s all it took to be able to start testing the new AI capabilities.
We chose a few stories from one of our recent software releases and used DevOps Aviator to create test scenarios. To conduct the test, we pulled user stories into ValueEdge and converted them to features. From there, we asked DevOps Aviator to create test cases for us. The DevOps Aviator function is really simple to use and fits perfectly into our existing workflow; essentially there’s a button in ValueEdge that says, “This is my feature, generate the test cases for me.”
Proving the concept
We started our testing by having both DevOps Aviator and our manual testers write test cases for Pick n Pay’s online shop and our mobile app. We then compared their output, and very quickly saw that DevOps Aviator was absolutely spot on. On average, 8 out of 10 of the DevOps Aviator suggestions matched the output from our manual testers. In addition, there was always at least one valid test case that the manual testers had overlooked. DevOps Aviator also added 20% coverage on our testing for platform specifics. So, right away, we saw that DevOps Aviator could not only do the job, but actually add value.
At that point, fully comfortable that DevOps Aviator was in line with the type of test cases we write, we stopped writing test cases and instead asked DevOps Aviator to show us the test cases we needed to run for each feature. We went through this testing process for about 45 features, which has given us complete confidence in DevOps Aviator’s ability to give us the results we need.
Faster, easier, more comprehensive
In addition to improving consistency and standardization between manual testers, we can see that DevOps Aviator will optimize our staffing resources by freeing testers from the routine writing of test cases and enabling them to focus on the more complex aspects of their jobs, such as exploratory testing and scenario analysis. DevOps Aviator also allows for faster feedback loops, enabling us to provide scenarios upfront to testers in sprint planning.
By allowing us to automate earlier in the process, DevOps Aviator will save effort, accelerate testing, and enable us to have a lot more coverage. When it comes to software test automation, you normally wait for the manual testers to complete at least one round of testing, then put the feature into production, and then automate. But with DevOps Aviator we can change the process, introducing automation even while the team is still busy developing.
We’ll no longer need to spend two or three days writing test cases—that will completely disappear from the equation. Our standard goal has been to have between 75% and 80% of new features automated for QA testing or system integration testing. With DevOps Aviator, we’ll be able to increase that to about 95%, as a result of eliminating those two or three days of waiting for manual test cases to be written.
As a further benefit, we can see that DevOps Aviator will help us onboard junior resources faster. Our model in Pick n Pay is to grow our teams from the junior level, so we have an intern program for new testers. We think that DevOps Aviator will shorten the onboarding process by at least six weeks by providing suggestions for the new staff to evaluate.
We’re now eagerly looking forward to DevOps Aviator graduating out of beta, since we can see that it will save us a lot of time and effort, as well as accelerate our software lifecycle and help us get new talent up to speed. This is something we’re definitely interested in adding to our DevOps toolset!
Hear more about how Pick n Pay plans to take advantage of AI when I take part in the keynote presentations at OpenText World Europe 2024 the week of April 15 in London, Paris, and virtually.
Guest author: Leon Van Niekerk is Head of Testing at Pick n Pay Group, a South African retailer. The company operates three brands-–Pick n Pay, Boxer and TM Supermarkets. Pick n Pay also operates one of the largest online grocery platforms in sub-Saharan Africa.
The post How Pick n Pay plans to transform software test cases with AI appeared first on ONE OpenText Blogs.