One of the key steps in any incident management process is creating a list of potential fixes to resolve the problem at hand. This process can be time consuming, and if the problem is severe, minutes matter. bigAI's suggestions can help you start finding solutions faster and reducing your time to resolve.

Issue Resolutions

To see suggested resolutions for Issues, go to the Resolve tab. This will trigger suggestion generation which may take a few moments to complete.

Inputs included in generation

Data Diagnosis

If row-level data access has been enabled in your Workspace, then bigAI will be able to query the underlying data in the table the issue occurred on in order to identify anomalous patterns and outliers.

The underlying data is not stored or persisted anywhere in Bigeye, nor is it used to improve or enhance the models used by bigAI.

Past Issues

bigAI can analyze prior issues that have been resolved, and understand if past suggested resolutions were used to successfully resolve the issue.

Incident Resolutions

Marking Suggestions as Worked or Didn't Work

Suggestions can be marked as "Worked" or "Didn't Work" to indicate to bigAi if a suggested helped you solve the Issue or Incident. This feedback will be used to make future suggestions more accurate.

This feedback is not used to retrain or tune models, it is passed in as context in future calls made to the bigAI model.

Generating More Suggestions

Click "More Suggestions" to have bigAI generate additional suggestions. These will not replace the existing suggestions but will be appended to the bottom of the list.

What Does bigAI Analyze?

bigAI is agentic in that it can take multiple steps in sequence in order to generate the final set of suggestions.

1. Basic metadata about the issue: anomaly size, time started, duration open, assignee, etc.
2. Analysis of underlying data: bigAI will execute 1-2 queries (no more than this) on the underlying dataset and look for patterns or outliers in the results.
3. Upstream ETL jobs: bigAI will review the code of the upstream ETL job and identify potential issues like incorrect join conditions, lack of filters to prevent nulls or disallowed values, etc.