AI-Driven Support Intelligence (Prototype)

How to run

python src/generate_raw_dataset.py
python src/export_raw_to_sheet_files.py
python src/process_raw_threads.py

Problem

Internal data platforms generate high-volume Slack support threads across multiple channels (Workbench, Pipelines, Query Builder, Data Quality, Tableau). Support leaders need quick visibility into: issue types, response time, resolution time, and escalation risk — without manually reading every thread.

Approach

This repo demonstrates a minimal “Raw → Processed” pipeline:

1. Synthetic raw Slack threads generated into:

• data/raw/raw_support_threads.json (source of truth)

2. Processing script produces:

• data/processed/processed_support_threads.json/.csv/.xlsx

The processing adds:

• issue_type (LLM or deterministic MOCK mode)
• thread_summary
• response + resolution metrics in minutes