# Databox Generative AI

Tracking a metric tells you what happened; deciding whether that's good, bad, or worth acting on takes more effort — usually pulling up historical data, comparing periods, and spotting what changed. Databox's Generative AI does that interpretation for you: it reads the historical data behind your metrics and writes a plain-language summary of what changed, along with recommendations for what to do next, wherever a performance summary is available.

## What generative AI does for you

Instead of digging into raw data to understand performance, generative AI surfaces the interpretation directly:

- **Understand performance without digging into raw data** — a written summary replaces manual comparison across time periods.
- **Spot trends and anomalies faster** — meaningful changes and outliers are called out instead of buried in a chart.
- **Make smarter decisions with less effort** — recommendations point to specific areas worth acting on, based on the data analyzed.


## How insights are generated

Powered by OpenAI, Databox's Generative AI turns raw historical data into a summary in three stages:

1. **Data preparation** — Databox pulls the relevant metric's history, including the comparison period needed to measure change, and cleans it for consistency.
2. **Pattern detection** — the model analyzes that data to identify the overall trend, meaningful changes, and any outliers worth flagging.
3. **Plain-language output** — the analysis is converted into a short written summary plus targeted recommendations, so the interpretation is ready to read rather than something you have to do yourself.


Each generated summary is made up of a performance indicator, a summary paragraph, and up to three recommendations — see [Review AI-powered performance summaries](/review-ai-powered-performance-summaries) for what each part means and how to regenerate them.

## Where you'll see generative AI in Databox

- **Hovering over a dashboard or metric** shows a performance indicator; clicking it generates or opens a summary. See [Review AI-powered performance summaries](/review-ai-powered-performance-summaries) for details.
- **Adding the AI Summary Datablock to a dashboard** generates a similar summary for up to 10 metrics at once, and lets you edit the result afterward. See [Understanding the AI Summary visualization](/understanding-the-ai-summary-visualization).


## What to keep in mind when using AI-generated insights

### Use AI as a starting point, not the final answer

AI can surface useful trends and suggestions, but it doesn't replace human judgment. Always review recommendations in the context of your business before acting on them.

### Context matters

Insights are based on the metric data available to Databox and may not reflect factors it can't see, like a seasonal promotion, an internal project, or a shift in the market.

### Watch for anomalies

A single spike or outlier can skew a summary's framing. If a summary seems off, check the underlying data source or try a different timeframe before trusting the conclusion.

### Be aware of potential bias

Because the underlying model learns from historical data, it can carry over patterns and assumptions from that data, including bias. Use the insights to guide your thinking, not dictate it.

No. Performance summaries and the AI Summary visualization are governed by their own daily and monthly generation limits, separate from the AI credit pool that powers [Genie, routines, and MCP](/understanding-ai-credits). See [Review AI-powered performance summaries](/review-ai-powered-performance-summaries#summary-generation-limits) for the specific limits.

Generative AI here is passive — it writes a summary automatically wherever one is available. [Genie](/get-started-with-genie-the-databox-ai-assistant) is a conversational AI assistant you actively ask questions, and can do more than summarize, such as building artifacts, dashboards, or creating metrics.