AI Disclaimer

Overview

Second uses artificial intelligence to help businesses process orders, parse documents, and manage operational workflows. While our AI systems are designed to be accurate and reliable, AI technology has inherent limitations that users should understand.

This disclaimer explains those limitations and sets expectations for how Second's AI-generated outputs should be used.

AI Can Make Mistakes

In an effort to be a helpful and efficient tool, Second's AI can occasionally produce outputs that are incorrect or incomplete. This is a known limitation of current generative AI technology, sometimes referred to as “hallucination.”

For example, Second's AI may:

• Misread or misinterpret data from emails, PDFs, or other documents
• Extract incorrect quantities, SKUs, pricing, or delivery dates from orders
• Fail to recognize certain document formats or unusual order structures
• Produce structured data that appears correct but contains errors
• Miss relevant information or context from source documents

Always Verify AI Outputs

Second is designed to handle the heavy lifting of order processing, but human judgment remains an important part of any AI-assisted workflow. We recommend:

• Maintaining human oversight of AI-processed workflows, especially for high-value or unusual orders
• Establishing internal processes for spot-checking AI outputs on a regular basis
• Using Second's confidence signals and flagging tools to prioritize which orders deserve a closer look, rather than reviewing every item manually

Why Mistakes Happen

AI limitations can stem from a variety of factors, including:

• Ambiguous source data: Poorly formatted emails, handwritten documents, or inconsistent order templates can be difficult for AI to parse accurately
• Novel formats: Orders or documents that differ significantly from patterns the AI has encountered before may be processed less reliably
• Context gaps: AI may lack the business-specific context needed to interpret certain abbreviations, product codes, or industry terminology
• Model limitations: The underlying AI models have inherent constraints in reasoning, arithmetic, and handling edge cases

What We Do to Improve Accuracy

While no AI system is perfect, we are committed to continuously improving the accuracy and reliability of Second's outputs:

• We use multiple validation layers and cross-checks in our processing pipeline
• We continuously refine our AI models based on real-world performance
• We provide tools for users to flag and correct errors, which helps improve future accuracy
• We design our interface to make it easy to review and verify AI-generated data against source documents

No Warranty of Accuracy

Second does not warrant or guarantee the accuracy, completeness, or reliability of any AI-generated output. All AI-processed data is provided on an “as is” basis.

Users assume responsibility for verifying AI outputs and for any business decisions made based on data processed by Second. For full details on liability limitations, please refer to our Terms of Service.

Reporting Issues

If you encounter an error or unexpected behavior in Second's AI processing, we encourage you to let us know. Your feedback directly helps us improve accuracy for all users.