AI Isn’t Replacing Testers — It’s Helping Us Do Better Work

 QonfX was my first in-person conference experience and I wasn't quite sure what to expect going into it. The theme — “The Future of Testing,” with a focus on AI in QA — was right up my alley. 

I'm always on the lookout for ways to improve and streamline processes. I figured at the very least, I could gain some insights into what other QA leaders are doing to introduce AI into their workflows, and what I can take back to my own team.

And ok maybe I was a little starstruck to see Michael Bolton was going to be present. 

My favourite session was Lavanya Mohan’s experiment — building and testing an app entirely with AI tools. She used GitHub Copilot to scaffold a simple app, but most of the heavy lifting of testing came from prompting ChatGPT. The importance of getting the prompt just right was evident in a few comical ways.

When she asked ChatGPT to generate 30 unique pizza toppings, it started strong — but by topping #20, it began repeating itself. Classic AI quirk. 

Upon further investigation, it was noted that the prompt had correctly asked for 30 distinct toppings, but the uniqueness was only enforced on the IDs, not the names. A great reminder of how precision in prompts really matters.

The first step in the testing was to set up the page object models, which she did by providing the HTML source directly in the prompt. This was the HTML generated by CoPilot previously. In reviewing the locators, she noticed they were not using Playwright best practices. 

She thought something wrong with the prompt, but after some digging realized it was actually an issue with the app itself - she had never had CoPilot build the app with suitable locators. This is an issue to be mindful of in developing with AI.

The work seemed to go a fair bit smoother on the API testing side. ChatGPT was able to add some methods to the POM and the endpoints tests cases were generated without much fanfare.

I do wonder — with all the back-and-forth checking of AI output — did it actually save time in the end? I didn’t get to ask her, but I’m planning to run my own experiment and find out.

All the speakers tended to allude to and agree on one fact: AI is not coming for our jobs, but it is changing how we work. This isn't necessarily a bad thing - quite the contrary. 

But as seen above, it can get you in a lot of trouble very quickly if you don't know what you are looking for. Therefore, you need to know good testing in order to spot bad AI.

📌 “When working with AI, humans are at the core and need to be the drivers.”

That line really stuck with me. It’s a reminder that AI isn’t the expert — we are. It can help us move faster, but it still needs our judgment to move in the right direction.

So with my team of very talented testers, I hope we can put our heads together and build a prompt that can help us run our Jira tickets through the AI to generate a list of test cases. It's a small step, but has the potential for big impact. If we can even reduce our time spent on test cases by 50%, it frees us up for more in-depth and creative exploratory testing - the kind of stuff we love.

All in all, my first foray into the conference world? A total success. I met so many fantastic QAs, and left feeling inspired. It's quite eye opening to realize that despite how varied our products are, we all boil down to the same fundamentals of software testing - it's mind blowing how many ways that can be applied!

Always curious to hear how others are incorporating AI in their workflows - let's chat :)

Comments

Post a Comment

Popular posts from this blog

Understanding HTTP Response Status Codes

Image
As a quality assurance professional, understanding HTTP response codes is essential, especially when testing APIs or web applications . These codes provide vital information about the status of requests, helping you identify issues quickly. In this post, we’ll break down what each response code means, why it matters, and how you can use it for effective testing. HTTP response codes are divided into five categories: informational, success, redirection, client error, and server error. While there are many codes in each category, the most common ones you'll encounter are 200 , 404 , and 500 . 1xx – Informational Responses You’ll rarely encounter these during typical testing, unless you're working with proxy servers or handling long-running requests . These responses simply indicate that the server has received the request and is processing it. 100 : Continue — The server has received the request headers, and the client should proceed with sending the request body. 101 : Switchi...
Read more

API Testing Overview

Image
When you're new to API testing, it can seem like a very foreign concept. In reality, it is very similar to testing any application. We'll look at some general categories to cover, and what types of tests those might include. Functionality Functionality tests are of course a major part of any test plan. In API functionality testing we check for things like a correct Response, the correct Response Code, and Schema Validation. You'll want to validate these for each endpoint you're working with, as well as for each type of request available. Security For the most part, APIs are typically internal tools and you'll want to protect the data from being accessed by outsiders. While security testing APIs we check for authentication and authorization methods, unauthorized access, and injection attacks. Performance At the most basic level, performance testing an API involves checking that the response time in within limits. This can be further fleshed out while the API is under...
Read more

Testing vs. Checking

Image
In software quality assurance, the terms testing and checking are often used interchangeably, but they actually refer to two distinct activities. While both are essential in ensuring the quality of a product, they serve different purposes. Let’s break down the differences and explore some common methods used for each. What’s the Difference? Testing is an in-depth process used to explore and assess the behavior of a system under various conditions. It’s about learning and discovering how the product performs, often revealing unexpected issues or areas for improvement. Checking , on the other hand, is a more straightforward process. It focuses on confirming that the system meets predefined expectations or requirements. It’s about verifying whether specific functions or features are working as they should. Key Differences Purpose: Testing: Aims to identify problems by exploring the system in various scenarios. It’s used to find defects, unexpected behaviors, or gaps in functionali...
Read more