Case Study · 03
AI-Assisted Engineering Operations
Operational conversations turned into an automated pipeline and retrievable engineering knowledge.
Operations ran on chat. Literally.
Operational signals arrived as messages in a WhatsApp group. Someone had to watch the channel, read each message, judge whether it mattered, chase context, and start a resolution. When an issue was resolved, the knowledge of how it was fixed lived in the conversation — and left when the conversation scrolled away.
Impact · AI-Assisted Operations
Turning operational conversations into engineering knowledge.
Engineering operations transformation through AI-assisted automation and knowledge reuse — not an AI product. Operational signals that once lived in chat now flow through an automated pipeline and return as retrievable engineering knowledge.
100%
Manual WhatsApp monitoring eliminated
Engineers no longer watch chat channels for operational signals — the pipeline does.
~90%
Engineering workload focused on technical engineering
The operational loop that consumed attention now runs automatically.
Signal
WhatsAppOperational messagesAutomation
n8nAutomation workflowUnderstanding
Vertex AIUnderstanding & triageState
PostgreSQLStructured state- ResolutionAction taken
Memory
QdrantVector knowledgeReuse
Knowledge RetrievalAnswers for future issuesReuse
Future issuesRetrieved answers short-circuit re-investigation
An automated pipeline with a memory.
This was an engineering-operations transformation through AI-assisted automation, not an AI product: operational messages now enter an automated workflow that understands them, records state, and routes resolution — and the outcome of every resolution is captured as retrievable knowledge for the next time the same shape of problem appears.
How it works
- Operational messages arrive in WhatsApp and enter an n8n workflow.
- Vertex AI interprets the message: what broke, where, and how urgent.
- PostgreSQL records structured operational state and history.
- Resolution actions are executed and confirmed back to the channel.
- Each resolution is embedded into Qdrant as vector knowledge.
- Future issues retrieve past resolutions before a human re-derives them.
From channel-watching to engineering.
The people who once watched a chat channel now do engineering. The knowledge that once evaporated into conversation history now returns answers to future incidents — the operational loop improved itself instead of repeating.
Outcome
- 100% of manual WhatsApp monitoring eliminated.
- ~90% of engineering workload focused on technical engineering.
- Resolved issues become retrievable knowledge for future ones.
- Operational attention redirected from watching to building.