J.S.Cock · AI Inquiry Demo
← All inquiries · Inquiry detail view
End-to-end flow

JT124028

One inquiry, all stages. Each box below is the actual artifact persisted by the pipeline — not a mockup. The brief asks: "does Kine actually work?". This page is the answer.

4 stages
  1. 1

    Email arrives in M-Files inbox

    Step 01

    Inquiry stored at .tmp/jscock_mfiles/JT124028/inquiry.eml

    ok
    From:
    Subject:
    Body: 0 chars
  2. 2

    AI extracts structured JSON

    Step 02

    0 item(s) detected · language=no · intent=new_inquiry

    ok
    Norwegiannew inquiry0 item(s)
    Show extracted JSON
    {
  "language": "no",
  "intent": "new_inquiry",
  "customer": {
    "company": null,
    "contact_name": null,
    "contact_email": null,
    "phone": null,
    "vat_number": null,
    "address": null
  },
  "items": [],
  "application_context": null,
  "attachments_referenced": [],
  "delivery": {
    "location": null,
    "deadline": null,
    "incoterms": null
  },
  "previous_reference": {
    "is_update_of": null,
    "notes": null
  },
  "completeness": "unusable",
  "missing": [
    "entire customer email — only the issued PDF quote was archived, not the original RFQ"
  ]
}
  3. 3

    Completeness verdict: unusable

    Step 03

    Missing: entire customer email — only the issued PDF quote was archived, not the original RFQ

    ok
    unusable
    • entire customer email — only the issued PDF quote was archived, not the original RFQ
    Rule trace (deterministic)
    Verdict: unusable · no items extracted
  4. 4

    Route to human

    Step 04

    Inquiry too sparse — sales engineer should handle manually.

    skip

Process map: see how this aligns with JS Cock’s diagram →