Generative AI For Testers

Gen AI is currently the buzz word in IT industry. Research and advancements keep happening in IT and learning never ends. Gen AI is an essential skill for career advancement and staying relevant in the industry. Let us start with some basic questions.

What is an LLM?

LLMs are Large Language Models. They can generate human like texts. There are many LLMs available and here are a few to mention, Chat GPT, Gemini, Anthropic, Open AI, DeepSeek, Llama.

How does an LLM work?

LLMs are trained on massive data sets of texts and uses neural networks that enables the model to handle sequences of data like sentences or lines of code. The neural architechture helps the models understand the context of each word in a senetence by considering it in relation to every other word. Then this architechture is trained in large amount of data. During training, the model learns to predict the next word in a sentence and then the model goes through several iterations, by which it learns how to predict the right word appropriate to the context. The pre trained data will then be fine tuned on specialized data.

Business Applications of LLM

Few of many uses of LLMs can be,

1.     Creating chatbots for customer queries, freeing up human agents for more complex issues

2.     Content creation like, articles, emails, social media posts or even video scriots

3.     LLMs can even contribute to software development and Testing, by generating or reviewing code.

As LLMs continue to evolve, we are bound to discover more innovative applications.

Gen AI

LLMs have certain limitations like, they rely on pre trained data and lack real time knowledge. They generate texts but can not perform tasks.

On the other hand, AI Agent uses artificial intelligence to perform tasks autonomously. It can make decisions and take actions to achieve specfic goals. AI Agents combines LLMs with other tools like APIs and decision making algorithms.

Gen AI in Testing

1. AI agents can be connected with existing test automation frameworks(e.g., Selenium, Playwright, Cypress) for UI testing

2. It can be integrated with API testing tools for backend validation

3. It can also be linked with CI/CD pipelines for automated triggering of test runs

Prompt Engineering

Prompt engineering is a process of designing and refining task specific prompts. It is a skill set that testers will have to learn when working with AI agents for testing. The way prompt is designed is directly related to the output that AI agent will generate. Lets take a quick look into the basics of prompt engineering.

Prompt Elements

·       Instruction - It is the specific task that we want the AI agent to perform

·       Context - Expains the details that helps the AI agent understand the task better

·       Input Data - It is the test data that is provided to the AI agent . Example, username and passwords in a log in test or a sample code that needs to be tested out in a code editor of an application

·       Output Indicator - This part includes the instruction of the format in which we like to view the test report

 Prompting Techniques

·       zero shot prompting - The AI agent depends on the pre trained data and no examples of any input data or output indicator will be provided by the tester

·       One shot prompting - Single sample data set will be provided

·       Few shots prompting - Multiple examples will be provided to AI agent in the prompt

Hope this blog post provides a basic insight about role of Gen AI in software testing. There is always more to learn. Lets keep learning!

Thank you!