AI analysis

IoT Pulse includes an AI analysis agent — IoTA — that reviews your devices, spots anomalies and developing problems, and raises tickets when something needs attention. The AI Agent screen is where you see what it found, review its findings, and act on them.

Open the AI Agent screen

In the sidebar, open IoTA → AI Agent. It’s available to operators as well as admins — you don’t need an admin role to use it. The screen has three views you switch between with the control at the top:

Command Center — fleet health at a glance, plus the findings that need you.
Triage Inbox — an email-style queue for reviewing findings one at a time.
Activity Feed — a timeline of every analysis the agent ran.

Command Center — what needs your attention

The Command Center is a triage dashboard. Across the top, five cards summarize the agent’s recent work:

Needs review — insights awaiting your validation.
Critical & high — high-severity findings across the fleet.
Analyses today — how many analyses ran recently.
Avg confidence — the agent’s average confidence in its open insights.
Agent spend — what the agent has cost this month.

Below the cards, the Needs your attention panel lists the insights waiting on you. Each card shows the device, a severity badge, the finding, and a confidence bar. For a finding you haven’t reviewed yet, use the quick actions:

Validate — confirm the finding is real.
Ticket — turn it into a maintenance ticket.

Once you’ve reviewed a card it shows Validated or Dismissed instead. A side rail shows Today’s runs (runs, insights found, tokens, cost) and an Agent activity feed of the most recent analyses.

Triage Inbox — review findings one by one

The Triage Inbox is an email-style review queue. Scan findings in the list on the left, and read and act on each one on the right.

Search the list, or filter it with the chips: All, Needs review, Critical, High, Anomalies.
The detail pane shows What IoTA observed, any Recommended actions, and a link to the Source analysis that produced the finding.
Act on the finding with the buttons at the bottom:
- Mark as valid — the finding is real and worth acting on.
- Mark as invalid — dismiss a false positive.
- Create ticket — open a maintenance ticket from the finding.

Activity Feed — the timeline of every analysis

The Activity Feed is a chronological story of every analysis the agent ran, grouped by day. Filter by analysis type (Device performance, Fleet health, Environmental) or show All activity.

Expand any run to see the agent’s summary, the insights it produced (each one links through to its detail), and the tools it used along the way (shown as small pills). Runs still in progress show as Running, so you can watch a live analysis as it works.

Run analysis on demand

You don’t have to wait for the next scheduled run:

On a device detail screen, select Run analysis to analyze that device now (see Devices).
On a device group detail screen, select Run analysis to analyze the whole group (see Device groups & shifts).

The Run analysis button at the top of the AI Agent screen points you to these per-device and per-group actions.

Each device also surfaces its most recent finding in a Last AI Assessment card on the device detail screen, so you can see the latest verdict without leaving the device.

Control how often analysis runs

How frequently the agent reviews a device is driven by a few settings, from most specific to least:

Per device — a device’s Criticality (Low / Normal / High / Critical) and its Analysis Interval (minutes). Mark important equipment higher so it’s reviewed more often (see Devices).
Per device group — a group’s Criticality and Analysis Interval apply to all its devices (see Device groups & shifts).
Per organization — tenant-level defaults set the baseline cadence when a device or group hasn’t overridden it (see Org admin).

See all agent runs in one timeline

The Agent activity screen (also in the IoTA section) is a unified timeline of every Analysis, Chat, and Simulation run across your organization — useful when you want analysis runs next to everything else the platform is doing. Filter by agent type and status, pick a date range, and select a run to expand its details. A summary strip shows the completion rate, cost, token usage, and run count for the range you’re viewing.