AI Guidance

Note: This document reflects the current “beta” state and will be updated as the feature continues to evolve. Please note that UI/UX elements may slightly change by the final release.

AI Guidance, powered by Sembi iQ, accelerates the initial selection of parameters and values, a crucial first step in the modeling process. This proves particularly valuable for newcomers to the test design approach, as well as more experienced users dealing with extensive documentation about complex systems under test.

Available AI Actions

AI Guidance is available for five specific actions within the Parameters section, exclusively in the Standard view. This feature is not accessible when working in Bulk view.

1. Suggest initial parameters: In a new, empty model, click the "Generate Parameters" button located below "Add Parameter" and "See Example Model" to create a set of initial parameters.
   
2. Additional parameters for an existing model: Click the "Generate Parameters" button at the end of your current parameter list to expand your model with additional parameters.
   
3. Generate values for a new parameter: When creating a new parameter with uncertain values, use the “Generate parameter variations” link in the "New Parameter" dialog to suggest appropriate values.
   
   Action #3 will NOT be shown if the AI Guidance modal is missing a goal or an upload (see more in the next section below).
4. Expand parameter values: Hover over any existing parameter name and select "Suggest more parameter values" to increase the selection of test values.
   
5. Reduce parameter values: Hover over any existing parameter name and select "Suggest fewer parameter values" to streamline and focus the test scope.
   

The availability of actions #4 and #5 adapts dynamically based on how many values your parameter currently has. The breakpoint is set at 5 values. Parameters with 5 values or fewer will default to expanding suggestions, while those with 6 or more values will focus on reduction.

Controlling who has access to the AI Guidance

If you are an Organization Admin on designwise.ranorex.com, you can open the “Organization Settings” under your profile name, then click the green button in the top-left corner, and click “Edit Organization” in the pop-up. Now, in addition to “Email domain list”, you have the checkbox at the top to control the overall AI availability for your organization.

    

You can set the default file expiration (“0” means files do not expire) and the max one. Users cannot exceed that max value in their model settings, but they can set a value lower than the default as well as enable expiration when the default is “0”.

When the “Enable AI suggestions” box is checked in the overall Organization settings, the column below becomes editable in the User Management view. There, you can enable or disable individual accounts based on your preferences.

If you are on a dedicated instance with the DesignWise+ licensing (i.e., companyname.ranorex.com), please reach out to your customer success manager to adjust the AI enablement and the file expiration settings.

How It Works

The process is flexible & iterative, as is the modeling workflow in our tool. Whether you are starting with the AI-generated draft or the manually-created model, you can leverage the different AI actions and consider adjustments to the goal & system documentation to continue refining your model until it is ready for testing.

The typical, initial set of steps could follow the pattern below.

Stage 1 - Context Setup

Create a new, empty model and click the “Generate Parameters” button.
To use AI Guidance, you must provide 2 essential pieces of context for the AI to work with:

• A high-level testing goal of at least 10 words that clearly articulate the scope and objectives of your modeling effort;
• Relevant documentation about the system under test in supported formats, with each file not exceeding 50 MB.

Once you click “Yes, provide guidance”, you will be redirected to the dedicated AI dialog.

Note: you can also get to that screen directly by clicking the icon in the left vertical menu or the one next to the model name
             

Until both parts are provided, you will see the red dot next to the AI icon at the top.

On the left, a user provides a high-level testing objective, describing the scope of what they aim to achieve with this model. For instance, if we were testing the Monopoly game implementation, the goal verbiage could be something like

“Ensure the system correctly implements the rules of monopoly. Focus heavily on the rules regarding money changing hands between players and player bankruptcy, and game-ending scenarios.”

On the right, a user adds the majority of their source documentation in a variety of formats. We strive to make this as flexible as possible so that a user doesn’t have to do extensive data cleaning or reformatting. This feature also expands our import flexibility beyond strictly formatted Excel files or mind maps.

With that said, the initial release will only support text-based documents - the AI extracts and processes only textual content from your documentation. When you upload files such as PDFs or PowerPoint presentations with both text and images, only the text portions will be analyzed. Files consisting entirely of non-text content, like images (JPG) or videos (MP4), will be ignored by the system.

Special mention - .opml mind maps are currently not processed by AI, you can either import them directly in the model creation dialog or convert them to e.g. .xml for AI.

You can provide relevant documentation in two ways. Upload allows you to submit files you have already prepared. Type/Paste lets you directly input information that our system compiles into a file for you. The AI processes both methods identically. The sequence in which you upload multiple files doesn't affect the AI's guidance. You cannot download attachments (uploaded by you or other users).

Lastly, you have the option to configure automatic file deletion after a specified timeframe. Note that this setting only applies to future uploads—it doesn't affect files already added to your model. If you upload a file with expiration disabled and later want to enable it, you'll need to manually delete the existing file and reupload it with the expiration setting activated. As mentioned above, this option respects the Organization-level settings. To remove the file manually, hover over it and click the trash icon.

Note: for more information about the data handling in our AI implementation, please see the “Essential FAQs” section.

Your final screen at this stage could look something like this:

Stage 2 - Generate Initial Draft and Review

With both context elements in place, we can close the Guidance dialog via the “Confirm” button and click the “Generate Parameters” button from Action #1 again to initiate the AI process. 

Once it finishes, a user will see the suggested set of key model elements - parameters and values - in a familiar UI setting. 

While it looks very similar to our Parameters screen, think of it as a “subspace” for review purposes only. You cannot directly edit parameters or values, and if you try to navigate away, you will see the following warning:

Note: while regenerating the suggestions is fast & easy, AI results are generally not 100% consistent, so it may be better to finish the review before doing anything else in the model.

There are a couple of things to highlight for this stage:

• The AI is “aware” of such techniques as equivalence classes and boundary testing that are connected to our Ranged Values and Value Expansions, so they can be added to the suggestions when the AI “deems it appropriate”.
   
• You can reject individual values, accept/reject the whole parameter, and accept/reject the entire set via the bulk options at the bottom.
  
• Only when all the review decisions are made, the screen will change to the Standard Parameters view that allows you to perform any typical modeling actions like editing parameters/values, navigating to Rules or Scenarios, etc.
  

Stage 3 - Refine and Iterate as Needed

AI Guidance operates as an intelligent assistant, presenting suggestions for your consideration, rather than making any immediate changes to your model. You maintain full control, reviewing each suggestion and deciding whether to accept or reject it based on your expertise and testing requirements.

Once you're satisfied with the AI's suggestions, you don't need to repeatedly specify the Testing Goal and System Documentation for subsequent AI actions. The AI also maintains awareness of your model's current state, automatically considering any manual changes you make when generating future suggestions.

When initial suggestions miss the mark, or you need a different focus for subsequent actions, simply return to the AI Guidance dialog to modify your testing goal or system documentation. This flexibility ensures the AI stays aligned with your evolving needs. Action #2 ("Additional parameters for an existing model") offers an additional dialog where you can specify a more targeted coverage goal to guide the AI's exploration of specific requirement areas. For instance, in our monopoly example, that additional goal could be “Test the scenarios for a player getting out of jail”.

Essential FAQs

• For a quick refresher without leaving the AI Guidance modal, open the Usage section in the top right.
  
   
• We recommend uploading files first, then specifying the testing goal (because the files take a bit of extra time to process, even after the upload spinner has disappeared). You may encounter a 15-second timeout message if the files require more time.
   
• Each “Generate Parameters” action will most often return no more than 10 parameter candidates. If you are expecting a larger model, you would need to trigger the same action a couple of times (you can use the “Pre-fill” option from the extra dialog above to automatically copy the goal).
   
• For more complex models, it can be a good idea to intentionally “feed” the context to AI in an incremental manner. Start with the core goal and the most important documents, and do the initial generation in the empty model. Make manual corrections, if needed. Then upload a few more documents and utilize the extra goal dialog to re-focus AI on the deeper exploration of specific elements.
  
  While the goals you provide are not used directly as prompts (we do the prompt engineering in the background so you don’t have to worry about that), the approach would be conceptually similar to chain prompting.
   
• Adding more specific wording to the testing goal (e.g., “telehealth scheduling scenarios”) can help the LLM pull extra aspects from the foundational knowledge pool, but it can also introduce the elements that are too generic or invalid/irrelevant for your particular scope. Experimentation is required to find the right balance in the goal wording based on your use cases.
   
• Text formatting is not considered for AI processing (e.g., a crossed-out paragraph would still be analyzed).
   
• AI Guidance can be enabled or disabled at the user level per request - please reach out to support or your company representative. It is enabled by default for the trial accounts.
   
• AI Guidance API has a limit of total requests on a monthly basis, 250 per user account (each AI action = 1 request; it resets on the 25th of each month).
   
• For future releases, we are evaluating the AI applicability to Scripts and Rules as well. For this one, to avoid any doubts, your interactions with other screens (Rules, Scenarios, etc.) don’t change.
   
• At least in the initial release, we will not be offering additional LLM training or fine-tuning options. I.e. uploaded system documentation is not used to change/update the LLM behind the scenes, it is only used as context for parameter & value generation. However, AI is receiving the initial goal and documents as well as the latest model content for each request, so it has the “ability” to gradually make better suggestions.

For more information, please refer to our AI Data Usage Policy.