Loading_
Loading_
A full ITIL v4 service management suite with AI triage, replacing a 9-year-old ticketing stack for a 14,000-seat manufacturer.
Measured outcomes
11s
Median first response
−68%
Mean time to resolve
4.8/5
End-user CSAT
$2.4M
Annual run-rate saving
The incumbent tool had become a queue, not a service. Tickets arrived unclassified, sat unassigned for an average of 41 minutes, and 34% were routed to the wrong team at least once. Change management lived in a spreadsheet.
We rebuilt the practice end to end: a classification model trained on four years of resolved tickets, a live CMDB fed by discovery rather than manual entry, and remediation playbooks wired directly to the incident record so common failures close themselves.
Headline result
0%
tickets auto-resolved
Tags
Five capabilities that define the system. Each one exists because something specific was broken.
Every inbound ticket is categorised, prioritised, deduplicated against open incidents and routed in under a second, with a confidence score that gates auto-action.
When a matched playbook exists, the platform executes it, verifies the fix, and closes the record with a full audit trail — no human touch.
Proposed changes are scored against blast radius from the CMDB graph, historical failure rate and current freeze windows before reaching CAB.
Agents and agentless scans reconcile into a dependency graph, so impact analysis reflects reality rather than last year’s documentation.
Breach risk is forecast from queue depth, assignee load and historical handling time — escalation happens before the clock runs out.
Event-driven core with a strict separation between the record of truth (Postgres), the decision layer (inference workers) and the action layer (sandboxed runners).
4 components
One catalogue rendered across four channels; the bot and the portal share a schema.
4 components
Inference is stateless and horizontally scaled; every verdict is logged with its inputs.
4 components
Runners execute in ephemeral containers with scoped credentials and hard timeouts.
4 components
Ticket state is transactional; search and analytics read from projections.
No mystery components. Everything below is either open source or a platform you already own.
Interactive mock-ups of the shipped interface. The live environment is available during a demo session.
Live queue with AI confidence badges and SLA heat
The running environment is available during a booked session — including a sandbox tenant you can drive yourself.
The real sequence, in order. Steps with a command are copy-pasteable.
Terraform module lays down AKS, Postgres Flexible Server and the key vault.
$terraform apply -var-file=prod.tfvarsHelm chart with per-environment values and sealed secrets.
$helm upgrade --install itsm aiinfraengine/itsm -f values.prod.yamlPoint AD, Intune, vCenter and the monitoring stack at the discovery ingest endpoint.
Backfill historical tickets, shadow-run classification for two weeks, then flip intake.
Published rather than hidden behind a call. Volume and multi-year terms move these numbers.
Platform licence
$18per agent / month
Implementation
from $85kone-time
Need this scoped against your estate? We will size it properly, in writing, within a week.
Request a quoteWe will walk you through the architecture, the trade-offs we made, and what would change for your environment.