This document outlines the limitations and appropriate use of the AI-powered PDF conversion system integrated into L2L Guides.
AI System as a Support Tool
The AI conversion system is an intelligent assistant designed to streamline the process of converting PDF documents into the L2L Guides platform. It is not a substitute for human judgment, expertise, or responsibility. Its primary purpose is to accelerate content creation and support our users, not to replace manual review and refinement.
Potential for Errors and Inaccuracies
The AI system is a complex model that analyzes document structure, text, and images to generate work instructions. While it is highly advanced, it is not infallible. It may produce content that is:
- Inaccurate or Incomplete: The AI's interpretation of PDF content may miss critical details, misread text (especially in scanned documents or images), or fail to extract all relevant information from complex layouts.
- Incorrectly Structured: The AI may misinterpret the hierarchy of content, merge unrelated sections, or split content inappropriately when creating guides, topics, and instructions.
- Based on Misinterpretation: The AI may misunderstand technical terminology, safety warnings, procedural steps, or visual content, leading to incorrect or potentially unsafe instructions.
- Missing Context: The AI lacks domain expertise and may not understand industry-specific requirements, safety protocols, or regulatory compliance needs that should be included in work instructions.
User Responsibility
As a user of this system, you are responsible for critically reviewing and validating all AI-generated content. This includes, but is not limited to:
- Thorough Content Review: Always review the entire converted content for accuracy, completeness, and correctness before publishing or sharing instructions with end users.
- Verification of Critical Information: Cross-reference technical specifications, safety warnings, measurements, part numbers, and procedural steps with the original source documents and subject matter experts.
- Document Verification: Ensure the correct documents are uploaded to prevent sharing sensitive or confidential information. Double-check file selection before submission.
- Quality Assurance: Ensure that images are correctly extracted and positioned, steps are in the proper sequence, and all safety-critical information is accurately represented.
- Exercising Professional Judgment: The AI's output should be used as a draft or starting point. Your expertise and knowledge of your operations are essential for creating safe, accurate, and effective work instructions.
- Reporting Issues: If you identify significant errors, systematic problems, or misleading conversions, please report them immediately to L2L Guides support. Your feedback is vital for improving the system.
AI Generation Disclaimer
Please treat all AI-generated content as a preliminary draft. While this system helps accelerate document creation, it is provided "as-is" and requires human verification. To prevent errors or safety discrepancies, please ensure a qualified individual reviews the output for accuracy and completeness prior to use.
Your Data is Safe and Secure
We take your data privacy seriously.
How We Protect Your Data:
- Your files are sent for processing over secure, encrypted connections (HTTPS)
- Only authenticated users with proper permissions can use the AI conversion feature
- Converted content is saved directly to your L2L Guides platform
- Access controls ensure only authorized users can view and edit converted content
- Uploaded PDF files are stored by Google for up to 48 hours and then automatically deleted.
The file is not "stored" in the traditional sense (like a cloud drive for long-term keeping). Instead, it is uploaded to a temporary "inference environment." Because AI models cannot read files directly from a local computer, the PDF must be uploaded to Google's secure environment, where it is "tokenized" (converted into a format the AI can understand). The 48-hour window is a maximum limit designed for efficiency. It allows us to ask multiple questions about a document without re-uploading and re-processing it for every single query.
The files are encrypted and linked solely to the specific API Key used by our application. The files are not accessible via public URLs.
What Gets Analyzed:
- Only the PDF files you explicitly upload for conversion
- Document content is sent to the Gemini AI service within our organization’s Google Workspace for processing.
- Text, images, and document structure are analyzed to generate L2L Guides instructions
What We Don't Do:
- We don't share your documents with third parties beyond Google Gemini for processing.
- Your data remains private and is excluded from Google's model training processes.
- We don't use your documents or converted content to train AI models
L2L Support can control access to this feature through permission settings and disable this feature at any time if requested.
FAQs
- Can you share which specific model and version of Google Gemini are you using for this functionality?
Answer: To ensure the highest quality of output and security, our engineering team maintains a rigorous update schedule. As Google releases performance improvements and patches to the Gemini model, we update our integration to match. Consequently, the specific version number changes frequently to ensure you are always leveraging the most capable and secure model available. The current product is using gemini-2.5-flash.
- Can you share where/in what Countries is inference environment for processing of pdfs located
Answer: The inference environment for processing PDFs is co-located with your existing L2L environment.
When the processing service is calling the Gemini API, these requests go to Google’s nearest available inference infrastructure by default.
- Can you share what safeguards are there against leaking of PDFs from inference environment and how do you make sure they are deleted in 48h
Answer: L2L maintains a paid enterprise subscription with Google for the use of Gemini. Unlike the free consumer versions of these tools, our commercial agreement explicitly prohibits Google from using any customer data to train their foundation models. Your data remains private to your instance, is used solely for the requested analysis, and is excluded from any Google learning processes.
The L2L processing service has been updated to trigger immediate file deletion upon the completion of analysis, even in the event of an error.
- Is there any possibility of uploading other files than PDFs, even though tool is theoretically made only for PDFs (underlying technology is General Purpose AI after all)
Answer: While the underlying Gemini technology is a general-purpose AI capable of processing various inputs, the L2L platform enforces strict application-level controls. Our software is hard-coded to accept and process only PDF files. Attempting to upload other file types will result in immediate rejection by the system, ensuring the tool is used only for its intended document processing purpose.
- Can you elaborate on following points: Your data remains private and is excluded from Google's model training processes.
See answer #3.