Skip to content
MLNetworks
Talk to an Expert

AI Agents

It reads the alarms. You decide.

AI Agents are the assistants in MLNetworks that read an incident and its alarms and write a summary that marks what was measured and what was inferred, for a person to review.

An agent reads the alarms on an incident, writes up what it found and says what it could not prove. It proposes a next step. A person reviews it and decides.

Talk to an Expert
Incident QueueINC-2041
Confirmed by the correlation agent

Summary in words

Preliminary, may change

Written by AI from the correlated alarms

Probable cause

Generator did not start during a grid outage

Confidence in this wording: medium

Key points

  • Grid lost 7 times in 24 hoursmeasured
  • No generator run recorded on the fifth outagemeasured
  • Battery discharge follows each outageinferred

From alarms to a reviewed decision

  1. 01

    The agent reads the record

    It works from the alarms, the ranked candidate causes and the correlation steps already on the incident. You can see what it read.

    Incident QueueINC-2041

    INC-2041

    Ridgeview RDG-0417

    Confirmed by the correlation agent

    The record this summarises

    • Correlated alarms

      3 alarms, one incident

      3

    • Candidate ranking

      Generator ranked first

      2

    • Correlation steps

      Grouped by site and time

      4

    • Impact

      Site on battery, generator not running

      1

    The answers this text paraphrases, as they were recorded.

    Site power issue

    • Mains failureFirst to raise
    • Generator failed to startSymptom
    • Low battery voltageSymptom

  2. 02

    A summary, each point tagged

    The summary is labelled as written by AI. Each key point is marked measured or inferred. It lists what is unproven and what to check next.

    Incident QueueINC-2041

    Summary in words

    Preliminary, may change

    Written by AI from the correlated alarms

    Probable cause

    Generator did not start during a grid outage

    Confidence in this wording: medium

    Key points

    • Grid lost 7 times in 24 hoursmeasured
    • No generator run recorded on the fifth outagemeasured
    • Battery discharge follows each outageinferred

    What is unproven

    Why the generator did not start.

    What to check next

    • Generator start circuit
    • Fuel level at the time of the fifth outage

  3. 03

    A proposal waits for a person

    The proposed action is held. A person gives a reason, approves or rejects, and the decision stays on the incident.

    Incident QueueINC-2041

    Proposed repair actions

    Inspect generator start circuit

    Waiting for a person

    Reason *

    No start on the fifth outage. Send the field team with the evidence attached.

    RejectApprove

    Actions taken

    • 03:20Summary written by AI, marked preliminary
    • 03:41Proposal reviewed by a person
    • 03:42Approved, with reason

Agents read what your systems already log

Integration scope is defined for each deployment, and each source is validated with your team before it is named as supported.

  • Alarm and fault systems
  • Site controllers and telemetry
  • Site and asset records
  • Work orders and tickets

Made to be checked

  • Confidence

    Preliminary, may change

    Confidence in this wording: medium

    Confidence in the wording, not in the cause itself.

    Confidence, stated plainly

    The summary states how confident the wording is. That is not confidence in the cause. While the alarms are still moving it is marked preliminary.

  • Agent revisions

    • Not confirmedThe correlation agent has not confirmed this incident
    • ConfirmedConfirmed by the correlation agent
    • WithdrawnShown if the agent withdraws the incident

    Revisions and withdrawals are visible

    If the correlation agent revises its view or withdraws an incident, the record shows it, with the current state.

The decision stays on the incident

What the agent prepared sits on the incident, with the reason and the person who decided. The field visit starts from that record.

Network Monitoring & Incident Response

Decision on the record

Inspect generator start circuit

Approved

Reason

No start on the fifth outage. Send the field team with the evidence attached.

Prepared by the agent. Decided by a person.

A person stays in charge

  • Agents investigate and propose

    They read the incident, explain what they found and suggest a next step. They do not act on the network.

  • Nothing hidden

    AI-written text is labelled, and each point says whether it was measured or inferred.

  • Every decision has a reason

    Approve or reject, with a reason. It is kept with the incident.

FAQ

Questions from operators

Do agents act on the network by themselves?

No. Agents investigate, explain and propose. A person reviews each proposal and approves or rejects it, with a reason.

How do I know what the agent actually saw?

The summary sits beside the record it paraphrases: the correlated alarms, the candidate ranking and the correlation steps. If the two differ, the record is right.

What if the agent is unsure?

It says so. The summary states its confidence, can be marked preliminary, and lists what is unproven and what to check next.

Which systems can agents read from?

That is defined for each deployment. We validate each source with your team before we name it as supported.

Can an agent be wrong?

Yes. That is why the reasoning is shown and a person decides. Summaries can be rewritten, and revisions stay visible.

AI Agents

Which investigation would you like prepared?

Bring a real case. An MLNetworks expert will walk through it with you.

Talk to an Expert

More of the platform

Every part uses the same sites and assets. Start where the need is greatest and add the next when you are ready.