DiDi AI by Findd

How we use AI

• DiDi is Findd’s conversational AI assistant designed specifically for blue-collar workforce operations. DiDi helps employees and managers get quick answers, manage time and labor tasks, receive important alerts, and access company information—all through natural language via SMS and email. DiDi is built to simplify day-to-day workforce interactions so employees can focus on their work, not on navigating complex systems.
  
  DiDi uses AI to understand messages from employees and managers, find the right information, and respond in a helpful and conversational way. When someone sends a question or request, DiDi’s central orchestrator routes the message to the appropriate specialized agent. The following sections describe the primary AI-driven capabilities covered in this disclosure.

Knowledge Base

• The Knowledge Base feature allows employees to ask questions in plain language and receive answers sourced directly from their employer’s own documents—such as employee handbooks, HR policy manuals, safety guidelines, and site-specific procedures.
• Important: The Knowledge Base only surfaces information from documents the employer has uploaded and made available. DiDi does not generate answers from external sources or make up information. If no relevant document is found, DiDi will let the employee know it cannot answer the question.
• How it works:
  •When an employer uploads documents to Findd, the system breaks them into smaller sections (“chunks”) and stores them using a technology called vector search.
  •When an employee asks a question (e.g., “What is the PTO policy?”), DiDi compares the question against all available document chunks to find the most relevant matches.
  •DiDi then uses a large language model (LLM) to compose a clear, natural-language answer based on those matched sections.
  •Documents are organized at both the account level (company-wide policies) and the employee level (site-specific or role-specific documents), so answers are tailored to the individual’s context.

Time & Labor Management (TLM) Exceptions

• DiDi’s TLM capabilities help managers stay on top of timecard issues and labor compliance. The AI analyzes timecard data to identify exceptions—such as missed punches, overtime approaching thresholds, or schedule deviations—and surfaces them proactively.
• How it works:
  •DiDi monitors timecard records and applies configurable rules to detect anomalies (e.g., an employee approaching 40 hours before the week ends, a missing clock-out, or a punch that doesn’t match an assigned schedule).
  •Managers receive plain-language summaries of exceptions and can ask DiDi follow-up questions like “What jobs are over budget on hours this week?” to get immediate, data-driven answers. Important: DiDi identifies and reports TLM exceptions for manager review. It does not automatically approve, reject, or modify timecards. All decisions regarding time and pay remain with the employer’s authorized personnel.
  •Escalation alerts that notify supervisors or management when issues require attention based on configurable severity criteria
• Alerts are delivered via SMS, email, or in-app notifications depending on the employer’s configuration and the employee’s communication preferences. DiDi’s AI is used to compose alert messages in clear, natural language and to determine the appropriate routing and priority of each notification.

AI and Decisions

• DiDi does not make employee-related decisions. DiDi is designed to provide information, surface data, and deliver notifications to support human decision-makers. It does not autonomously approve or deny leave requests, hire or terminate employees, set compensation, or take any binding employment action.
  
  When DiDi provides recommendations—such as flagging a timecard exception or suggesting that a compliance issue needs review—these are presented as recommendations for a qualified human to evaluate. The final decision always rests with the employer’s authorized managers or HR personnel.

Training Data & Architecture

• DiDi uses a retrieval-augmented generation (RAG) architecture. This means DiDi does not rely solely on a pre-trained AI model’s general knowledge. Instead, it retrieves specific, relevant information from the employer’s own documents and data before generating a response.
• • Foundation Models: DiDi is powered by large language models (LLMs) that have been trained on broad datasets of publicly available text. These models provide DiDi’s ability to understand and generate natural language.
  • Client-Specific Data: Each employer’s documents (handbooks, policies, procedures) are processed and stored securely using vector embeddings. This data is used at query time to retrieve relevant context—it is not used to retrain the underlying LLM.
  •Policy Documents: Employer-configurable policy documents guide DiDi’s behavior for each account, ensuring responses and actions align with the specific employer’s rules and preferences.
  • Data Isolation: Each employer’s data is logically isolated. DiDi never shares one employer’s documents, employee information, or configuration with another employer.

Testing & Monitoring

• Findd maintains rigorous testing and monitoring practices to ensure DiDi performs accurately and reliably:
  • Quality Assurance (QA): DiDi undergoes automated and manual QA testing, including scripted test scenarios for each agent type (knowledge base queries, TLM lookups, alert delivery, and HR communications). QA scripts validate that responses are accurate, relevant, and appropriately sourced.
  • Data Governance: Findd follows strict data governance practices. Employer data is encrypted in transit and at rest, access is controlled by User ID lists per account, and data is never used to train third-party models.
• • SOC 2 Certification: Findd has achieved SOC 2 certification, demonstrating compliance with industry-standard security and data protection controls. • Transparency: Employees interacting with DiDi are informed they are communicating with an AI assistant. DiDi does not impersonate a human. • Equitable Access: DiDi’s multilingual translation capabilities help ensure that non-English-speaking employees have equal access to important workplace information and communications.

Questions & Contact

• We are committed to being transparent about how AI is used in our products and welcome any feedback or concerns. If you have questions about how Findd uses AI, please contact us.