---
title: "wattiot | Smart analytics: anomalies and AI agents"
description: "wattiot learns each machine's normal consumption, spots what leaves that range, and an AI agent has the report ready before the shift starts."
source: "https://wattiot.io/en/use-cases/analytics/"
language: "en"
---

# Smart analytics

> wattiot learns each machine's normal consumption, spots what leaves that range, and an AI agent has the report ready before the shift starts.

**Use case · Analytics**

wattiot learns what is normal for each machine, spots what leaves that range, and an agent has the report ready before the shift starts.

## The situation: Nobody looks at the data at three in the morning.

Compressors do not stop when production does. A leak, a valve left open or a badly scheduled dryer keeps drawing power all night, and on the dashboard it looks like any other night.

Seeing it takes knowing what is normal for each machine at each hour, and someone checking every day. In a small plant, that someone does not exist.

| With wattiot | Today, without wattiot |
| --- | --- |
| What is normal for each machine, learned from its own data. | Fixed thresholds nobody updates. |
| What leaves the range, with the report already written. | Someone has to catch it in time. |
| Your questions, in plain language. | Export to Excel and cross-check by hand. |

## The schema: From the compressor room to your inbox.

The same signals that already reach wattiot Core feed anomaly detection, the agents and the assistant. There is no second system to integrate.

1. **Compressors** — Power of the compressors and the dryer, and air flow, over Modbus TCP. → _Modbus TCP (reads)_
2. **wattiot Edge** — Reads, stores locally if the network drops and sends encrypted to the cloud. → _TLS 1.3 (encrypted stream)_
3. **wattiot Core** — Learns each signal's normal band and flags whatever leaves it. → _03:00 (every day)_
4. **AI agent** — Every day at 03:00 it reviews the night, writes the report and emails it.

## From data to result: From a normal band to a report in your inbox.

Nobody sets a threshold: the band comes from each signal's own history. When the night leaves it, the agent finds it, puts a number on it and raises it. Whatever is left to know, you ask the assistant.

1. **Normal band** — Each signal's usual range, P10–P90, learned from its own history.
2. **Anomaly** — Production stops at 22:00, but the air stays at 47 kW and leaves the band.
3. **Agent** — At 03:00 it reviews the shift, works out the kWh with no production and acts.
4. **Report and assistant** — The report lands in maintenance's inbox. For the rest, the assistant.

_Compressed-air consumption from 18:00 to 06:00 against its normal P10–P90 band: production stops at 22:00, but consumption stays around 47 kW, out of the band from 22:30, adding up to 213 kWh with no production by 03:00. At that hour the night-watch agent writes the shift report and emails maintenance; afterwards the assistant answers questions about the air's consumption. Sample data._

## Spec sheet

| Spec sheet |  |
| --- | --- |
| Sources | Two compressors, a dryer and an air flow meter |
| Protocol | Modbus TCP. Read-only |
| Detection | Normal P10–P90 band, learned per signal |
| Operational contexts | Production running or stopped, and shift |
| Agent | Temporal trigger at 03:00, Generar Informe and Notificar Email |
| Assistant | Plain-language questions about signals, contexts and KPIs |
| Alerts | Email, push and a shift report as PDF |
| On site | One wattiot Edge on the room's network |

## Other cases

- [Every kWh, with its order and its shift](https://wattiot.io/en/use-cases/energy/) — wattiot reads power from the line's OPC-UA server, totalises it into kWh and splits it across the orders and shifts that were running.
- [Every reel, with its whole history](https://wattiot.io/en/use-cases/traceability/) — wattiot reads from the PLC over EtherNet/IP which reel sits on each revolver axis and gathers its tension, length and passes into one record per reel.

## Does this look like your plant?

We will show you on a scenario like yours: what to connect first, what you will see and which plan fits.

[Book a demo](https://wattiot.io/en/demo)

---

- Equivalent HTML page: https://wattiot.io/en/use-cases/analytics/
- Agent index: https://wattiot.io/en/llms.txt · https://wattiot.io/en/agents.md
- Sitemap: https://wattiot.io/sitemap-index.xml