User Journey: Pipeline — Connect → Improve

The Command Center shows a pipeline strip with stages from Connect through Improve and Value. This page explains what each stage does and how they fit in the user journey.

Pipeline overview

The pipeline represents the end-to-end flow of data and operations in AtlasAI:


Connect → Discover → Reconcile → Visualize → Correlate → Diagnose → RCA → Act → Automation → Improve → Value

Left side (Connect → Visualize): Getting data in, discovering assets, reconciling with CMDB, and visualizing in dashboards.
Middle (Correlate → RCA): Turning events into incidents and diagnosing root cause.
Right side (Act → Value): Deciding and executing remediation, then measuring improvement and value.

Stage-by-stage

Connect

What it is: Ingest data from external systems and agents.

Where: Data Sources, Integrations (Prometheus, Datadog, PagerDuty, ServiceNow, Jira, Slack, AWS, Kubernetes, etc.), Edge Agent collectors.

How to use: Add connectors, configure credentials, and confirm data is flowing (metrics, logs, traces, events). The pipeline strip often shows counts (e.g. connectors, events received).

Outcome: Raw telemetry and events are in the platform.

Discover

What it is: Automatically discover services, hosts, and dependencies from the data you connected.

Where: Discovery jobs (often under Data Sources or Discovery), topology/CMDB ingestion.

How to use: Run discovery jobs (on a schedule or on-demand) so that services, pods, and relationships are populated. Discovery feeds the dependency graph and CMDB.

Outcome: Discovered entities and relationships available for topology and reconciliation.

Reconcile

What it is: Align discovered data with your CMDB and configuration records.

Where: CMDB, reconciliation jobs, topology views.

How to use: Map discovered entities to CI records, merge duplicates, and keep ownership and metadata in sync. Reconcile improves accuracy of blast radius and RCA.

Outcome: A single source of truth for assets and their relationships.

Visualize

What it is: Dashboards, service maps, and topology views.

Where: Dashboards, service map / topology, log and trace explorers.

How to use: Build dashboards and open topology to see health, metrics, and dependencies. Use filters (environment, service, time range) to focus.

Outcome: Operators can see system state and trends at a glance.

Correlate

What it is: Group related events and alerts into incidents (or attach to existing incidents).

Where: Correlation, correlation rules, event streams.

How to use: Define correlation rules (e.g. by service, time window, alert type). When events match, the platform creates or updates an incident instead of leaving many unrelated alerts.

Outcome: One incident per “thing that went wrong” instead of alert storms.

Diagnose

What it is: Triage and investigate incidents — assign, add evidence, set severity.

Where: Incidents, Command Center priority queue.

How to use: Open an incident, add evidence (metrics, logs, topology), assign to a team or person, and optionally run RCA from here.

Outcome: Incidents are triaged and ready for root cause analysis.

RCA

What it is: Root cause analysis — AI-reasoned hypotheses with evidence and suggested actions.

Where: RCA Lab, incident detail Run RCA.

How to use: From an incident, click Run RCA. The engine uses evidence, topology, and knowledge base to produce root cause and often a suggested runbook.

Outcome: Clear root cause and next step (e.g. run a specific runbook).

Act

What it is: Decide what to do — approve a runbook, execute manually, or let automation run.

Where: Incident detail (suggested runbook), Runbooks, Automation.

How to use: Review the suggested runbook; approve it (if in draft). Then execute from Automation (or use “Execute” from the incident). For L1/L2, you may approve step-by-step.

Outcome: A runbook is approved and ready to run.

Automation

What it is: Execute runbooks and automation jobs.

Where: Automation, runbook execution UI.

How to use: Start a run: select runbook, link incident, set variables. Monitor execution; if approval gates are on, approve steps. Execution logs and status are visible per step.

Outcome: Remediation runs (e.g. restart, scale, rollback); incident can auto-resolve when successful.

Improve

What it is: Learn from outcomes — update trust scores, tune policies, add runbooks.

Where: Automation history, trust/autonomy settings, runbook library, post-incident review.

How to use: Review automation success/failure; adjust autonomy level or approval gates. Add or refine runbooks for recurring issues. Use RCA and resolution data to improve future hypotheses.

Outcome: The system gets better over time (higher automation success, better RCA).

Value

What it is: Measure and report value — MTTR, MTTD, automation rate, cost savings.

Where: Value dashboard (if enabled), reports, FinOps, adoption metrics.

How to use: Open the value or adoption view to see KPIs (e.g. incidents auto-resolved, time saved, cost avoided). Use for stakeholder reporting and roadmap decisions.

Outcome: Clear proof of value and adoption.

How the stages connect

Connect and Discover feed Reconcile and Visualize (data and topology).
Visualize and Correlate feed Diagnose (you see issues and they become incidents).
Diagnose and RCA feed Act and Automation (you know what to do and run it).
Automation and Improve feed Value (outcomes and learning become metrics).

The Command Center pipeline strip is clickable: each stage links to the page where you do that work. Use it as a map for the full journey from data to value.