Silent Knowledge Extraction
“Your workers are already labeling your data — they just don't know it.”
📖 The Scene
At 06:42, Budi typed a quick message to his shift group about a noisy bearing. He didn't open a maintenance app or fill out a form. NayaCore silently extracted the asset, condition, and location from his casual chat, created an Evidence Card, and boosted the incident confidence to 83% after Eko corroborated it.
Budi
Machine Operator (T1)
📱 See It In Action
Real WhatsApp conversations between operators, supervisors, and NayaCore — in Bahasa Indonesia, exactly as they appear on the phone.
[Voice Note]11:15
Sterilizer dua agak lambat hari ini, biasanya 90 menit sekarang udah 100 menit belum selesai. Nggak tau kenapa.11:15
T1 Group Chat Silent Ingestion
NayaCore passively listens to field operator WhatsApp groups, extracting structured intelligence without disrupting the conversation.
Text Evidence Extraction
Operators discuss a strange noise at Press 3. NayaCore's NLP engine extracts the asset ('Press 3'), condition ('bunyi kasar', 'bearing kering'), and creates an Evidence Card behind the scenes. Eko corroborates, boosting the confidence to 83%.
Photo & Vision Processing
Rina posts a photo of a flooded area. NayaCore's Vision Pipeline runs object detection, while NLP parses the caption. It immediately cross-references this with Cakrawala weather data and flags a possible cross-module insight.
Local Voice Note Transcription
Joko sends a voice note. To protect PII, NayaCore runs a local Whisper model to transcribe the audio. It calculates a +11% cycle time increase and links it to MillOS sensor data.
📊 Before vs. After
Traditional Operations vs. NayaCore
| Metric | Traditional | NayaCore |
|---|---|---|
| Operator observations reaching system | 0% (data lost in chat) | >80% (auto-extracted as Evidence Cards) |
| Time from observation to logging | Never | < 5 seconds |
| Training required for operators | Hours (learning a new app) | $0 (they just use WhatsApp) |