Pilot candidateDocument workflows

RFP Response Generator

An AI-assisted document-response workflow built around evidence verification and controlled uploads.

Upload screen with two past winning proposals and a new RFP questionnaire attached, ready to generate a draft — Upload past proposals and the new RFP or questionnaire

The problem

Responding to RFPs is slow and error-prone — teams spend more time hunting for the right source information than writing, and unverified claims create real risk.

What it does

Upload 2–5 past winning proposals plus a new RFP or security questionnaire
Automatic requirement extraction from the uploaded RFP
Requirement-to-past-proposal matching with literal-excerpt citations
AI-drafted, cited proposal response exported as a Word document
Unsupported or invented claims are converted to explicit gaps instead of guessed answers
Password-gated production access with signed, expiring session cookies

Who it is for

Anyone assembling a cited response to an RFP or security questionnaire from past winning proposals.

Human controls

AI output always requires human review — pricing, legal promises, security claims, compliance claims, dates, and customer names must be verified before use. Model-returned quotes and citations must be literal excerpts from an uploaded file; anything that isn't becomes an explicit gap.

Verification

12 automated tests covering auth tokens, rate limits, same-origin checks, upload policy, and evidence/citation guards.

Limitations

Job store and rate limiter are process-local, suitable for one controlled instance rather than multi-instance public deployment; pipeline execution is fire-and-forget with no durable queue; scanned/image-only PDFs need OCR outside the app. The AI drafting pipeline has not yet been validated with a live model run; output quality and cost are unverified.

Stack

Next.js
TypeScript
Anthropic API
Word & PDF parsing

Updated

2026-09