The Challenge
A 70-year-old California custom manufacturer with 40,000 SKUs had a painfully manual quoting process. Custom specs flowed into a legacy software system, were manually re-entered into a second system, which then advanced the bill of materials and generated an RFQ response. The team consistently missed their 48-hour turnaround target and spent significant time on back-and-forth with customers to collect missing information.
What They Built
Remix Partners built an AI workbench inside Claude Code that converts legacy software output to machine-readable JSON, then automates the full RFQ workflow: catalog comparison across 40,000 SKUs, bill of materials generation, stakeholder notification, customer-ready pricing, and first-pass manufacturing drawing.
Remix Partners identified that the client's core obstacle wasn't ambition — they had already built Gemini-based internal tools — it was data access. The legacy software produced proprietary output formats that no downstream system could read. The first and most critical step was converting that output to JSON, which made 40,000 SKUs and decades of manufacturing data machine-readable for the first time. That single transformation unlocked a cascade of downstream automation. Remix Partners then built the AI workbench inside Claude Code over 3.5 weeks. The workbench receives an RFQ, compares it against the full catalog, generates a bill of materials, notifies the right internal stakeholders, converts the BOM to customer-ready pricing, and produces a first-pass manufacturing drawing — all before a human reviews it. The engineers who had been skeptical about AI's applicability to their specific process became the project's loudest advocates once they saw the prototype operating on their own data.
AI Role
RFQ turnaround time cut nearly in half from a baseline the team had routinely missed with their manual process
Infrastructure
• Claude Code (AI workbench environment) • Legacy manufacturing software (output converted to JSON) • Gemini (prior internal tool foundation) • Custom catalog and BOM processing logic
Integration Points
• Legacy software output converted to JSON via custom parsing layer • RFQ input connected to 40,000-SKU catalog comparison engine • BOM output linked to stakeholder notification and pricing conversion • Pricing output triggering first-pass manufacturing drawing generation
.JPG)
