feature · alerts

When the AI needs a human,
it picks up the phone.

Email, SMS, voice call, or a direct push to your existing systems — the channel is chosen based on how urgent the event is. A compressor idle for eleven minutes gets a phone call. A scheduled inspection miss gets a morning digest email.

/ how alerts are configured

No fixed tiers.
Configured per deployment.

Every deployment writes its own alert policy — what matters, what doesn't, who gets called, on which channel. Tell us in plain English at onboarding; change it the same way later.

Example policy · cold-store partner

“For compressor stalls over 10 minutes, call the maintenance lead. For dock doors stuck open over 20 minutes, SMS the dock supervisor. For safety incursions, call whoever is on the rota and try the next person if there's no answer. Everything else — just log it and put it in the morning digest.”

During the private beta

We configure it for you.

You give us the brief; our team writes and tunes the policy. Zero tooling on your side. Reviewed weekly with whoever owns the alerts on your team.

After the beta

You change it in plain English.

Self-serve in the Darkfield platform. Type the change you want — “only call me on stops longer than fifteen minutes” — and it's applied on the next alert cycle.

/ voice call

The AI on the phone.
This ships today.

Voice calls are a production feature, not a roadmap item. The AI synthesises a contextual greeting, selects the recipient from the on-call rota, places the call, and records the response. The call log lands in Dashboards as a structured event row.

HOW A P1 CALL WORKS
01
The AI detects the condition.
CAM-04 confirms compressor C-3 idle for 11 minutes. Detection precision on the machine state is 0.97; the AI escalates to P1.
02
Recipient selected from on-call rota.
The AI checks the current shift schedule. On-call is the maintenance lead for the cold-chain hall; that's who gets the call.
03
The AI calls and explains the situation in plain English.
No DTMF menus — the recipient just talks back. They can ask follow-up questions, request more detail, dismiss the alert, or escalate. The AI answers from the live event context and logs the conversation.
04
Conversation logged; event row created in Dashboards.
The call recording, transcript, and outcome land as a single structured row in Dashboards. If there's no response, the AI tries the next person on the rota.
Human Yes, hello.
AI Hi, this is your LineAlert AI Assistant. I'm calling with a Predictive Stoppage Alert. I can see that the Raw Prep station has not processed a new batch in the last fifteen minutes. Based on your current extraction speed at Station One, your Clean Buffer will be empty in twenty-two minutes.
Human Okay, can you walk me through the timings?
AI The Raw Prep station has not processed a new batch for fifteen minutes. At the current extraction speed at Station One, the Clean Buffer will be empty in twenty-two minutes.
Human Okay, what do we need to get everything in order?
AI There are two pallets of the necessary ingredients currently available in cold storage. It takes us twelve minutes on average to move, unload, and wash these ingredients. To avoid any delays, we need to initiate this process within the next ten minutes.
Human I see. Okay, thank you.

Just an example call.

/ escalation policy

Configured in plain language,
not a tree of dropdowns.

Escalation rotas are described at onboarding in natural language. "If the on-call doesn't acknowledge within ninety seconds, try the next person on the rota. If they don't respond in sixty seconds, send an SMS to the operations director." The AI parses this and creates the policy; you can edit it later in plain language via Dashboards.

EXAMPLE ESCALATION POLICY · cold-store partner
t=0 Voice call to on-call maintenance lead (from rota)
+90s No answer → AI tries the next person on the rota
+60s Still no answer → SMS to operations director
+∞ Event remains open in Dashboards until condition resolves

When something matters,
you'll know about it.

We're onboarding a small number of partners in private beta.

read about dashboards →