AI Usage Disclosure

1. Overview

RFP Pipeline uses artificial intelligence (AI) as a core component of its platform. This disclosure explains what AI models we use, how they process your data, and what safeguards are in place. We believe transparency about AI usage is essential, especially in the context of federal proposal preparation.

2. AI Models Used

RFP Pipeline currently uses the following AI models:

• Anthropic Claude: Used for proposal section drafting, content analysis, compliance extraction, opportunity classification, and source change detection
• OpenAI Embeddings: Used for semantic search within the content library to find relevant past content atoms

We may update the specific models used as improved versions become available. The fundamental architecture (isolated tenants, human-in-the-loop review, structured outputs) remains consistent regardless of the underlying model.

3. How AI Is Used

• Proposal Drafting: AI generates draft text for proposal sections based on your content library, the solicitation requirements, and the compliance matrix. All drafts require human review and acceptance.
• Compliance Extraction: AI reads solicitation documents to extract requirements, page limits, formatting rules, and evaluation criteria. Results are verified by our expert curator.
• Opportunity Scoring: AI assists in scoring opportunities against your company profile. Scores are calibrated by our expert curator and adjusted based on your feedback.
• Source Monitoring: AI classifies changes detected on federal procurement websites to determine if they represent new opportunities, amendments, or routine updates.
• Content Library Search: AI-generated embeddings enable semantic search to find relevant past content for new proposal sections.

4. Data Isolation

Each tenant receives isolated AI agent instances. Your data is never mixed with another tenant's data in AI context windows. AI agents provisioned for your workspace can only access: your content library, your proposal sections, your tenant profile, and the public solicitation data relevant to your proposals.

5. Data Handling by AI Providers

Our AI providers (Anthropic, OpenAI) process your data solely to generate responses to our API requests. Per our agreements with these providers:

• Your data is not used to train their AI models
• Your data is not retained beyond the API request lifecycle (typically seconds)
• Your data is transmitted over encrypted connections
• Your data is processed in the United States

6. Human-in-the-Loop

AI never submits content on your behalf. Every AI-generated output is presented as a draft that requires explicit human review and acceptance. The platform clearly marks AI-generated content with confidence scores and provenance tracking. You are solely responsible for reviewing, editing, and accepting any AI-generated content before including it in a proposal submission.

7. Limitations

AI-generated content may:

• Contain inaccuracies or fabricated information (commonly called "hallucinations")
• Produce text that does not accurately reflect your company's capabilities
• Miss nuances in solicitation requirements
• Generate content that requires significant revision

This is why human review is mandatory. The AI is a drafting assistant, not a proposal writer. Your expertise and judgment are the final quality gate.

8. Federal Compliance

Some federal agencies require or may require disclosure of AI assistance in proposal preparation. It is your responsibility to comply with any such requirements. RFP Pipeline can provide documentation of which sections received AI assistance upon request.

9. Questions

For questions about our AI usage, contact eric@rfppipeline.com.