import Cards from '@/components/docs/Cards.astro';
import Card from '@/components/docs/Card.astro';
import Alert from '@/components/docs/Alert.astro';

AI Analytics turns metering-node telemetry into ready-to-use reports for decision making. Instead of exporting device data by hand and assembling spreadsheets, you select a report, set its scope (company, node, period) and receive a finished document with figures, charts and conclusions.

## Deterministic core, AI as an explanation layer

Every report is built on a **deterministic analytical core**: each key conclusion — archive quality, period readiness for closure, data freshness, uncovered-volume estimate, Q/V reconciliation, sensor stability, link state, passport completeness, incidents and recommendations — has a formal origin, is computed by defined formulas and **reproduces exactly when re-run on the same data**.

AI is used as an **auxiliary language layer**. It briefly explains already-computed results in plain language, summarizes the deterministic indicators and produces readable commentary. It **does not** replace the formulas, establish facts, or decide statuses — final report statuses are determined by rules, thresholds and archive data.

<Alert type="warning" title="Core rule">
AI is an explanation layer, not a decision layer. Every status and recommendation carries its own explanation: which data were used, which formula was applied, which threshold fired and which limits apply to the conclusion.
</Alert>

## Two levels of reports

Reports answer at two levels. **Node-level** reports cover a single device or object — its consumption profile, battery life, pressure behavior. **Fleet-level** reports cover an entire fleet or a selected company — the most problematic nodes, the largest leaks, the economics of automation.

<Cards cols={2}>
  <Card
    icon="cpu"
    title="Node-level reports"
    description="Per-device analytics: consumption, battery, pressure, emissions, forecasts."
    href="/en/platform/v3/ai-analytics/node-reports"
  />
  <Card
    icon="network"
    title="Fleet-level reports"
    description="Fleet-wide analytics: problem nodes, leaks, failures, automation economics."
    href="/en/platform/v3/ai-analytics/fleet-reports"
  />
  <Card
    icon="ruler"
    title="Methodology"
    description="How the figures are computed and where the limits of trust lie."
    href="/en/platform/v3/ai-analytics/methodology"
  />
  <Card
    icon="sigma"
    title="Formula reference"
    description="Every metric and threshold on one page."
    href="/en/platform/v3/ai-analytics/formula-reference"
  />
</Cards>

## Trust and limits

Reports are a management and operational assessment for prioritizing decisions — **not an accounting or metrological act**. Every figure is tagged with a confidence level (confirmed, scenario or possible), so estimates are never passed off as measured facts. See the [methodology](/en/platform/v3/ai-analytics/methodology) for the full set of principles.