Artificial Intelligence

SpiraPlan's artificial intelligence functionality empowers you to automate the creation of essential project artifacts from requirements, such as user stories, features, epics, and business/system requirements. It allows you to quickly generate a set of standard test cases, tasks, risks, and BDD scenarios that can then be refined and improved as needed. We have versions that use both the public OpenAI GPT models and private Azure OpenAI LLMs.

Generative AI Using ChatGPT Plugin

The ChatGPT SpiraApp runs inside SpiraPlan and provides a set of dropdown menus with options to auto-generate derived SpiraPlan artifacts from the current requirement. For example, imagine that you have just created a new requirement or user story that consists of a single-line Ability to login to a web application or As a user I want to be able to login to a web application.

Normally, you would now need to manually write the various test cases that cover this requirement, including positive (can log in successfully) and negative tests (failure to log in for various reasons). In addition, you would need to decompose this requirement in a set of lower-level development tasks for the developers to create the user interface, database, and other items that need to be in place to have a working login page.

If using a BDD approach, you might also want to create a set of BDD Gherkin scenarios that describe each use case for a login page more specifically. Finally, you would want to identify and document all potential risks associated with this new feature.

Generating Test Cases

When you click on the option to generate test cases, the plugin will call the OpenAI API and create a set of test cases for the requirement in question. For our sample requirement, you can see that it has generated four test cases, one for the positive case and three additional negative cases:

Each of the test cases consists of a single test step that has the description of the test to be carried out, the expected result from the test, and any sample data that you might want to use:

Note that the sample data will most likely be very notional since it does not know valid/invalid logins for your application. Still, the overall structure is correct and will save a lot of manual time writing and documenting the test cases.

Generating Development Tasks

Next, we can click the generate tasks option to generate five standard development tasks typical for a web login page. You can see that they are relatively high-level but cover the key areas, both the front-end (creating pages) and the back-end (database, session management, authentication system):

Identifying Risks

Next, we would want to understand the risks of deploying a new login page. Since one of the most critical security attack vectors is an application's authentication and session management mechanisms, understanding these risks is very important and often overlooked. All you need to do is click the generate risks button, and the plugin does the rest:

In this example, you can see that it generated six different potential risks, including GDPR privacy risks, security vulnerabilities, performance issues, compatibility, and more. It also includes risks related to the user experience, which are also often forgotten until too late!

Generating BDD Scenarios

Finally, we can use the option to generate BDD scenarios to have the plugin create a set of positive and exception scenarios for using the login page:

In this example, it has created four scenarios, one positive and three negative. Each is written in the Gherkin Given... When.... Then, format and is ready to use.

Azure OpenAI Functionality for Azure Users

For customers in more regulated industries, there may be concern with sending their data to the public OpenAI API that is needed to generate the artifacts with ChatGPT. Although OpenAI's terms of service state that API queries are not used to train the underlying Large Language Model (LLM), we also have a more private solution available as well. SpiraPlan includes a separate Azure OpenAI version of the plugin that lets you connect to your private LLMs hosted inside Microsoft Azure, accessible via. the same REST API as the public ChatGPT version.

Since each instance of the OpenAI GPT models is hosted privately inside a customer's Azure environment, you can use the private URL for your own instance of Azure OpenAI. This ensures that you keep control over your sensitive data, while at the same time benefiting from the productivity improvements of Generative AI.