AI-Powered IT Event Management
Letting support staff ask questions in plain language instead of sifting through logs
The Challenge
A national retail chain's IT team was drowning in system events. When something broke, finding the right event data took too long. Support staff spent hours sifting through logs instead of solving problems. The volume and complexity of events kept growing. They needed a way to get to the answer fast.
They already had event infrastructure (EventArc). What they lacked was a way for people to interact with it without learning query languages or digging through raw data. They needed AI that could find the relevant events and summarize them so support could act.
The Approach
We built a system on Google Cloud's Vertex AI that let support staff ask questions in plain language. Type what you're looking for, get back relevant events, summaries, and next steps. The AI finds the right event data and summarizes it so people can act instead of hunt.
The system plugged into their existing EventArc infrastructure so we didn't rip and replace. We designed it to scale with their data volume. The architecture separated retrieval (finding the right events) from generation (summarizing and recommending) so each piece could be improved independently.
- Natural language querying. Support staff ask questions in plain English instead of writing queries
- AI-generated summaries and recommendations. Context-aware insights and suggested next steps for diagnosis and resolution
- Vertex AI architecture. Built on Google Cloud Platform, integrated with EventArc, designed for scale
The Results
Outcome: ~40% faster issue resolution; ~30% efficiency gains; staff can focus on fixes instead of log digging.
What They Got
Support staff could find the right events and act on them in a fraction of the time it used to take.
Less time hunting, more time fixing. The team could focus on solving problems instead of searching for clues.
- Faster identification of system issues through AI analysis
- Integration with existing EventArc infrastructure (no rip and replace)
- Architecture that scales with growing event volume
Technologies Used
Google Vertex AI, Google Gemini (large language model), Google Cloud Storage, BigQuery, EventArc. The system uses retrieval-augmented generation: it retrieves relevant event data and uses the LLM to summarize and recommend, so answers are grounded in actual events instead of made up.