When Oracle Analytics AI features find an anomaly, trend or key driver, let authors save those results as a dataset or subject area. Each run would add rows with the metric, the date found, the expected versus actual value, the severity, the main drivers and the filters applied. Authors could then build workbooks, schedule reports and set alerts on AI findings like any other data.
AI insights in Oracle Analytics are shown on screen at the time someone looks, and then they are gone. There is no history of what was flagged, when, or how often. That means we cannot:
- report on how many anomalies were found in a metric this quarter, or whether they are rising
- show a governance or board committee which issues AI raised and when
- combine AI findings with our own data, for example sickness absence spikes alongside staffing levels by site
- schedule or burst AI findings to the managers who need to act on them
Proposed solution
- A "Save insights as dataset" option on Explain, Auto Insights and anomaly visualisations.
- An option to rerun on a schedule and append new findings with a run date, so there is a history.
- Standard columns - metric, dimension values, date detected, expected value, actual value, variance, severity or confidence, top drivers, filters and data source used.
- The dataset respects the same row-level security as the source, so managers only see findings for their own area.
- Usable in workbooks, Publisher reports, alerts and data flows like any other dataset.
Business benefit
- Turns AI from a one-off on-screen feature into evidence that can be reported, trended & audited.
- Lets organisations prove the value of AI, for example - AI flagged 14 absence spikes this quarter and 11 were acted on.
- Gets AI findings to the people who need them through scheduled reports, not just to the author who happened to run Explain.
- Especially useful in regulated sectors such as healthcare, where early warning and an audit trail both matter.
