From requirements to test cases
Drop in a requirements document: the AI drafts test cases for each requirement, and every draft is compared with the document line by line. Nothing the document does not say goes into the results, and values the document does not state stay blank.
Contact usTo test requirements written in natural language, each one has to be turned into test cases with concrete conditions, inputs and expected results. Repeating that interpretation and writing for every requirement costs the engineer's time. With TC Gen, the AI writes the drafts and the engineer reviews and completes them.
How one requirement becomes test cases
Below is an example of test cases made from a single requirement sentence.
Put it in as it is, no tidying. Column names can differ from company to company.
When the wiper switch is AUTO and the rain sensor detects rain, the controller shall set WIPER_CMD to LOW.
- Test actionSet the wiper switch to AUTO and drive the rain signal RAIN_LEVEL to LIGHTExpected resultWIPER_CMD = LOW
The sentence says “set WIPER_CMD to LOW”, so the result is taken straight from the document.
- Test actionSet the wiper switch to AUTO and drive RAIN_LEVEL to NONEExpected resultExpected value unsettled
The requirement only covers rain being detected. It does not say what the wiper should do when RAIN_LEVEL is NONE, so no value is guessed and the cell stays blank.
Enlarged part of the screen. Click to open the full screen in a popup.
The actual results screen
Pick a requirement and the test cases made from it appear as a table.
Screen guide
- The list on the left is the requirements; the table on the right holds that requirement's test cases.
- Rows showing NOT SPECIFIED in the expected-result column are the blanks. The “expected value unsettled” mark in the list means the same.
- The Notes, Evidence and Source tabs show why each case was written that way.
Drafts that disagree with the document are dropped;
values the document does not state stay blank
As in the example above, each draft is handled according to how it compares with the requirements document.
Writes a draft test case for each requirement
Checks whether the draft's signals, values and conditions are actually in the document
- Disagrees with the documentDraft dropped
If the signals, values or conditions in a draft disagree with the requirements document, that draft is dropped from the results. The standard is agreement with the document, not how plausible the AI's text sounds.
- Document gives no resultExpected value unsettledOnly unsettled items need the engineer
If the requirement does not state the result for a particular condition, no value is guessed. The row is marked “expected value unsettled” and collected in the review list, so the engineer only completes those items. A value the engineer fills in is kept as a requirement note and carried into revisions, so the same blank does not come back.
- Passes the document checkIncluded in results
Test cases included in the results are linked to the requirement sentence and the reference-document location they were written from.
Gaps can be spotted and completed right away
Before analysis starts, the number of requirements read and the excluded rows are shown, so omissions surface before anything is generated. In the results, every requirement carries its test case count, so an empty requirement stands out, and any other document a requirement points to but cannot be found is flagged in the review list. The engineer only completes what is flagged.
Enlarged part of the screen. Click to open the full screen in a popup.
The evidence screen
Shows which part of the requirements document and reference documents each test case was written from.
Screen guide
- Items under a document name are the evidence found in that document. “Exact match” means the same name was found in the signal value table.
- Click an evidence item and the Source tab opens at that section of the document, with the sentence or table it found highlighted.
The work, screen by screen
Select a step to see the actual screen.
Enlarged part of the screen. Click to open the full screen in a popup.
Import the requirements document without editing it
Column names can differ from company to company. For columns that are not recognized automatically, you look at the document's column name and say whether it is the identifier, title, text or verification method. The same project uses this setting for later documents.
Screen guide
- The table at the top lists the column names read from the document. On the right you choose which requirement field each column is.
- Only the fields marked as required must be filled before moving on.
Connect the AI service you use, or a Local AI
The AI used for this work can be chosen per project. A Local AI is an AI running on your PC or a company server. Document files and generated results are stored on your PC, and the connected AI receives only requirement text and the reference content used for writing. The PopcornSAR server checks account and license only.
Files are stored on your PC
AI running on your PC or a company server. Works on a closed network too.
Only requirement text and reference content
AI service login or API key
Log in with your account or register an API key. TC Gen never stores or reads your login credentials.
TC Gen account and license only. No documents are sent.
With a Local AI, everything stays inside your network
- Data does not leave
- Requirement text and reference content are processed inside your network. Documents and result files are on your PC anyway.
- Works without internet
- Authenticate with an offline license file and run the whole flow, from import to review, on a closed network.
- The same safeguards
- Whatever model you use, the document comparison, the unsettled marks and the review list apply the same way. Even with a small model, values not in the document do not get into the results.
- Switch AI whenever you like
- Each project picks an AI service or a Local AI, and switching later does not change how documents are read and reviewed.
At a glance
- Requirements document formats
- docx, xlsx, pdf, pptx
- Reference documents
- Same file formats as the requirements document, plus signal value tables (dbc)
- Document languages
- Korean, English, Japanese
- Export
- xlsx, csv, json, using your company's column names
- Supported operating systems
- Windows, Linux, macOS
- License
- TC Gen account login, or a file for authentication without internet
Discuss adoption with the requirements document you use today
We discuss scope and schedule around the requirements document you use today.