Case study
RFP Responses Grounded in Approved Technical Knowledge
Approved company knowledge behind each technical response.
Previous approachA generic ChatGPT drafting workflow.
Purpose-built workflowRequirement-level drafts grounded in approved materials.
Technical documents and prior RFPs prepared for retrieval.
Relevant company passages used in requirement-level drafting.
Approximately 4 RFPs per day reported.
Engagement
Overview
Applied Optoelectronics needed a more company-specific approach to RFP preparation. The goal was to move beyond a generic ChatGPT workflow and make approved technical materials and prior RFP content available to the drafting process.
I built a retrieval-augmented generation workflow that connects document preparation to individual RFP requirements. Technical documents and scanned material pass through extraction and OCR, followed by normalization and chunking. The system retrieves relevant passages from that prepared knowledge to support each response draft.
The workflow separates knowledge preparation, retrieval, drafting, and review. A language-model endpoint supports response generation, while an explicit review/export gate keeps generated drafts within the proposal preparation process.
Usage figures are internally reported. Reporting periods and user definitions remain unconfirmed.
- Reported users
- 27
- Reported RFPs / day
- ≈4
Challenge
Useful RFP drafts needed the right company material for the specific requirement being answered.
Company grounding
Drafts needed to draw from approved technical knowledge rather than rely on a general-purpose conversation alone.
Document preparation
PDFs and scanned materials needed extraction before their content could support retrieval.
Requirement-level relevance
Different requirements needed different supporting passages.
Review control
Generated drafts needed a defined review step before export.
Solution
I connected the document pipeline to targeted retrieval and requirement-level drafting.
Extraction and OCR
Turn technical documents and scanned material into usable text.
Content preparation
Normalize and chunk extracted content into retrievable knowledge.
Targeted retrieval
Retrieve relevant passages for the requirement being answered.
Grounded generation
Supply the retrieved company content to the drafting workflow.
Review and export
Route generated responses through the documented review/export gate.
Retrieval at the level of the requirement
The workflow treats each requirement as a separate drafting task. Instead of relying on a single broad summary of company documents, it retrieves relevant technical passages for the question at hand. This connects the response-generation step to a focused set of company materials.
Impact
Reported usage shows recurring use of a document-grounded workflow in technical proposal preparation.
Reported adoption
27 users.
Reported activity
Approximately 4 RFPs per day.
Implemented capability
Extraction, OCR, technical-content retrieval, requirement-level drafting, and review/export.