Skip to content
Start free

Understand · AI Core

An AI that connects the signals — and shows its evidence.

AI Core reads across applications, network, assets and logs, groups related symptoms into one situation and tells the story in plain language. The readings behind every claim are one click away.

Start freeBook a demo
AI Core / Correlation
Example
Example: twelve alerts from network, application, asset and log monitoring collapse into one situation, BGP instability. Its Story reads: the BGP session from core-rtr-01 to AS64501 left Established at 14:02; interface errors, jitter on 12 devices and checkout-api latency rose in step; cause suspected — the flapping peer. Coverage is 8 of 9 sources, and the Evidence tab lists each reading against its expected range.

Overview

What is wrong right now, ranked.

Open AI Core and the first screen answers the first question of every shift: what needs attention, across every module, in order. Forecasts flag what is heading toward trouble before it gets there.

  • One ranked list across applications, network, infrastructure, databases and logs
  • Forecasts of what is heading toward trouble
  • Every finding opens in Ask with one click
AI Core / Overview
Example
Example: the AI Core overview ranks what is wrong right now across Network, APM, Databases, Assets and Logs; a flapping BGP peer on core-rtr-01 rises to the top as its score climbs to 92. A forecast warns that disk /var on db-01 is projected to be full in about 3 days.

Correlation

Forty alerts become one situation.

A BGP peer drops, interfaces start erroring, jitter climbs on a dozen devices. AI Core groups the related signals into a single situation with a plain-language Story, and lists the connected symptoms under “Also in this situation”.

  • Groups signals from applications, network, assets and logs
  • The Evidence view shows each reading, its expected range, and why it belongs
  • Handles BGP state changes, interface errors, ICMP jitter, failed power supplies, memory pressure, log-volume shifts and saturated services
AI Core / Correlation
Example
Example: the Evidence view of a BGP instability situation, cause labelled Suspected. Coverage reads 8 of 9 sources reporting, naming the missing one — syslog from dist-sw-03, not treated as zero. A chart of interface errors draws a missing interval as a hatched no-data gap, and each piece of evidence is labelled Observed, Suspected or Established.

Ask AI

Ask in plain English. Get an answer from live telemetry.

“Any network issues right now?” “Who is eating the bandwidth?” “Top 5 servers by CPU.” Ask answers from your own live data, with charts, tables and follow-up questions — not from a generic knowledge base.

  • Charts and tables in the answer, not just text
  • Suggested follow-up questions
  • Coverage stated in every answer, e.g. “8 of 9 sources reporting”
AI Core / Ask AI
Example

Who is eating the bandwidth?

7 of 7 flow exporters reporting

In the last 15 minutes the top talker on the core-sw-01 uplink is backup-01 (10.20.4.17) at 412 Mbps — 38% of the link, almost all of it SMB to nas-01. The next four talkers together use less than half that.

Top talkers · core-sw-01 uplink · Mbps
  • backup-01 → nas-01412
  • app-02 → db-0196
  • web-01 → cdn71
  • app-03 → redis38
  • ws-114 → updates22
Is this normal for this hour?Show the flow recordsWhich ports is it using?Evidence ›

Which services are slowest right now?

6 of 6 services reportingCause: suspected

checkout-api is the slowest: p95 1.84 s against a learned 420 ms for this hour of the week. Most of that time is spent waiting on payments-api POST /charge, which is also above its baseline.

p95 latency · ms · tick = learned baseline
  • checkout-api1,840
  • payments-api1,210
  • orders-api310
  • inventory140
  • gateway95
Show the slowest traceWhat changed at 14:02?Chart p95 for the last 6 hoursEvidence ›

Top 5 servers by CPU

8 of 9 sources reporting

db-01 leads at 91% average CPU over the last hour, followed by app-02 (78%) and app-03 (74%). web-04 has sent no readings since 13:40 — it is shown as missing, not as zero, and is not counted as healthy.

Average CPU · last hour · %
  • db-0191%
  • app-0278%
  • app-0374%
  • web-0252%
  • web-04no data
Chart CPU utilisationWhy is db-01 busy?When did web-04 stop reporting?Evidence ›
Example conversation with Ask AI using fictional data. Choose a tab to see a network, applications or servers question answered from live telemetry, with a chart, the coverage of the answer and suggested follow-up questions.

How it answers

Built to be trusted at 3 a.m.

An operator acting on an AI answer needs to know how sure it is. So AI Core is strict about what it claims: a missing reading is never treated as zero, silence is never reported as healthy, and every cause is labelled observed, suspected or established.

  • Answers state their coverage
  • Missing data is shown as missing
  • Causes are labelled, never overstated
  • The AI explains and suggests; people and approved automations act
Network Insights / Alert detail
Example
Example: on the alert “Device temperature high” for dist-sw-03 at 74°C, the AI Alert Insight node reads the temperature trend and fan speeds, then writes an automated comment: inlet temperature rose 9°C in 40 minutes while fan 2 reports 0 RPM; suspected cause fan 2 failure; next steps to check the fan, inspect airflow at rack B4 and replace the fan tray if confirmed. It suggests only; people and approved automations act.

Capabilities

Everything in AI Core

The complete list, as it exists in the product today.

Overview

  • What is wrong right now across every module, ranked
  • Forecasts of what is heading toward trouble

Correlation

  • Related signals from applications, network, assets and logs grouped into one situation
  • A plain-language Story for each situation
  • An Evidence view: the readings, the expected range, and why they belong together
  • “Also in this situation” shows the connected symptoms
  • Handles signals such as a BGP peer leaving Established, interface errors, ICMP jitter across many devices, a failed power supply, memory pressure across a group of servers, log-volume shifts and saturated services

Ask AI

  • Ask in plain English and get an answer drawn from your live telemetry
  • Charts, tables and follow-up questions in every answer
  • Open any finding in Ask with one click

How it answers

  • Answers state their coverage — e.g. “8 of 9 sources reporting”
  • A missing reading is never treated as zero
  • Silence is never reported as “healthy”
  • Causes are labelled observed, suspected or established — never overstated
  • Evidence is always one click away

AI across the product

  • AI Alert Insight node in Automations
  • Docs Assistant: explains, sets up, checks prerequisites, configures, verifies and troubleshoots each integration

Coming soon

  • Building dashboards by chatting with the AI
  • Bring-your-own LLM per tenant
  • AI-authored automations

On the roadmap, not yet available.

Works with AI Core

AI Core is where every module meets.

Each module on the platform feeds the same AI Core. The more sources you connect, the more of the story it can tell — and it always tells you which sources it heard from.

From AI Core

  • Applications: saturated services
  • Network: BGP peers leaving Established, interface errors, ICMP jitter
  • Assets: failed power supplies, memory pressure across a group of servers
  • Logs: log-volume shifts

Ask AI

  • “Any network issues right now?”
  • “Who is eating the bandwidth?”
  • “Top 5 servers by CPU”

Get started

See every signal in one place.

Start with one site and one probe. Add modules as you go — the AI Core gets more useful with every source you connect.

Sign upBook a demoRead the docs (opens in a new tab)

Already have an account? Sign in to the portal