Best Practices: Preparing AI-Friendly Input Documents for TestVista

Token Utility

Overview

TestVista works best when input documents are clean, structured, and consistent. This guide explains how to prepare your specification documents so TestVista’s AI can extract requirements and generate test cases effectively.

DOs and DON'Ts

Guideline Type	✅ DOs	❌ DON'Ts
Input File	TestVista supports Word, Markdown (supports both Basic Markdown and Extended/GFM Markdown), Excel and Image documents. Word format has the best generation quality.	Upload old format files such as .doc, .xls
Input Language	Ensure the document is written in a single language for the best results.	Mix multiple languages in one document.
Document Organization	Structure the document logically with clear headings and organized sections.	Use scattered or ambiguous formatting without clear structure.
	Provide sufficient context for AI to understand the requirement	Assume AI is magic and can understand domain knowledge
	Format requirements using structured frameworks like Given...When...Then such as User Story or Use Case	Use freeform or inconsistent requirement descriptions.
Input File Size	Keep files small and focused on one screen or feature. Input requirement < 10 pages works best.	Upload excessively large documents covering unrelated features.
Document Organization	Review and proofread the document to ensure there are no errors or inconsistencies.	Upload documents with typos, incomplete sentences, or unclear formatting.
AI guidance	Use detailed and explicit descriptions for complex requirements with many components.	Test the AI with overly complex requirement without explicit instructions.
AI guidance	For example: "Help me create test cases for all 20 components on login screen."
Probabilistic nature	Due to the probabilistic nature of AI generation, users can create different test suites and try regenerating the test cases of the uploaded document if the initial test cases do not meet quality expectations.	Rely solely on the first set of generated test cases without exploring different test suites or attempting regenerations to improve test quality and coverage.
Content	Keep only the core requirement content. Remove unnecessary, emotional, or historical info.	Keep business context or personal opinions, unrelated descriptions.

Below are some Best Practices for different types of documents.

1. High-Level Requirements (Usually in PowerPoint)

Preferred Format: Convert to .docx OR .md before uploading.

‍Why?
TestVista currently does not support PowerPoint format. Since PowerPoint slides are often visual and fragmented, converting them to Word ensures that the requirements are presented in a textual format that AI can process.

‍Best Practices:

Copy text from slides into a Word document, structured by headings and bullet points.
Add text description where the slides rely on visuals (e.g., diagrams, arrows).
Keep requirement statements clear and declarative

‍Example:

PowerPoint Slide	AI-Friendly .docx
Title: Login Module - Secure access for users	Heading 1: Login Module Requirement 1.1: The system shall allow users to log in using email and password. Requirement 1.2: Login must be secured with SSL encryption.

2. SRS (Software Requirement Specification) Documents

Preferred Format: Keep in .docx OR .md

Best Practices:‍

Use consistent heading levels (e.g., H1 for modules, H2 for submodules, H3 for requirements (H1 > H2 > H3)).
Number requirements (e.g., REQ-001, REQ-002) to avoid ambiguity.
Avoid mixing requirement statements with explanations in the same paragraph.

Example:

3. Detailed Design / Basic Design Documents

Preferred Format: Convert to .docx OR .md before uploading.

Best Practices:

Ensure diagrams (UML, sequence flows) are exported as images and inserted in Word.
Keep explanatory text below diagrams (not embedded inside shapes).
Use tables for structured information (e.g., API parameters, database fields).

Example:

Not AI-Friendly: Sequence diagram drawn in Word using shapes, with labels floating.
AI-Friendly: Export the diagram as a single PNG/JPG and insert it into Word, followed by descriptive text.

4. Documents in Excel (Spreadsheets) format

If your document is in Excel format, since Excel can be quite complex, here are some Best Practices to help AI read and understand it more accurately and effectively, such as:

One value per cell. Do not split a requirement across multiple merged cells.
Avoid merging cells, especially in wide columns (e.g., column AA, AB).
If using process flows or complex shapes, export them as images.

‍Example of Bad format:

How to improve formatting for better AI readability:

Issue (Bad Format)	Fixed (Good Format)
① Too many merged cells Merged cells make it hard for AI to identify individual data points.	One value per cell Keep each cell independent with a single value to ensure accurate data parsing.
② Shapes (arrows, frames, icons) AI cannot interpret shapes or extract meaningful content from them.	Replace with captured images Take a screenshot of the visual and insert it as a static image for better AI processing.
③ Floating textboxes used for notes/explanations Textboxes are not linked to the data and may be ignored by AI.	Write notes directly in cells Remove textboxes and place explanations directly below or beside the relevant data.

On this page

Overview DOs and DON'Ts 1. High-Level Requirements (Usually in PowerPoint)2. SRS (Software Requirement Specification) Documents 3. Detailed Design / Basic Design Documents 4. Documents in Excel (Spreadsheets) format