AI quote intake with vision and RAG for a lighting supplier

Custom system that turned 30–60 minute quotes into 5–10 minute quotes against a 1M+ SKU catalogue.

Dutch industrial lighting supplier selling into the trade. Catalogue: 1M+ SKUs across every category of professional lighting in the European market. Competes on depth and quote turnaround speed.

Problem

A single quote took 30–60 minutes of skilled human time. Inbound orders arrived as messy semi-structured documents: supplier-portal screenshots, Excel sheets, PDFs, photos of handwritten lists. Specialists read the email, interpreted the format, opened the catalogue, and searched line by line against 1M+ SKUs. No natural-language search. Two specialists, ten quotes per day, 20–40 hours per week of skilled labour on repetitive work. Output also varied between specialists.

How the industry typically approaches this

The same pattern shows up wherever high-volume unstructured documents hit a structured backend: a parsing layer, a matching layer, and a generation layer. The implementation has shifted from OCR plus rules to vision language models plus retrieval-augmented generation. Recruitment is the adjacent industry where this is most publicly documented.

Our approach

Custom AI quote intake sitting between the inbound email queue and the internal catalogue. Four stages. Ingestion: emails picked up automatically, attachments extracted, inputs passed into a vision-capable LLM that reads any format. Extraction: model identifies each line item (brand, SKU, EAN, quantity, spec notes) and flags ambiguities. Matching: extracted items matched against the 1M+ SKU catalogue using a retrieval layer that handles natural-language queries. Output: formatted quote ready for human sign-off. Dashboard shows quotes in flight and clarification loops.

• •Time per quote dropped from 30–60 minutes to 5–10 minutes (~80% reduction)
• •16–32 hours per week of skilled labour freed
• •Built in 10 days, running in production

What the industry has achieved with similar solutions

External benchmarks from comparable deployments. Sourced and labelled as third-party evidence, not our own results.

• High-growth HR tech client via 4Spot

  AI resume parsing into Keap CRM. 150+ hours per month of manual data entry removed. Single source of truth in the pipeline.

• Fortune 500 technology firm via MiHCM

  Enterprise AI recruitment. -40% agency spend (~$3.2M/yr). Time-to-hire 60 → 35 days. Recruiter productivity +45%. MiHCM