The AI Operations
Intelligence Brief

2026 is the year AI stopped suggesting and started doing. Agentic AI systems now plan, execute, and verify multi-step IT operations tasks — but autonomy without guardrails is an outage waiting to happen. Here's where agents genuinely help, and where humans must stay in the loop.

Digital twins have moved beyond manufacturing floors into the data centre and the NOC. Combined with spatial computing interfaces, they're changing how operations teams visualise infrastructure, rehearse changes, and predict failures before they happen.

Attackers are harvesting your encrypted traffic today to decrypt it when quantum computers mature. With NIST's post-quantum standards finalised and regulatory deadlines approaching, PQC migration has become an IT delivery programme — not a research topic. Here's the practical roadmap.

Most service desk teams are sitting on untapped potential. With the right prompt architecture, Claude or ChatGPT can triage tickets, generate incident summaries, draft RCA reports, and coach L1 agents — all without replacing your team.

How I built SmartDesk — an AI triage system that auto-classifies 42% of tickets — using n8n workflows and Claude API. A complete walkthrough of the architecture, prompt design, and lessons learned.

Not all organisations are ready for the same AI interventions. This maturity framework helps CIOs and service delivery leaders identify where they are — and what their next AI investment should be.

ITIL 4's Service Value System wasn't designed with LLMs in mind — but its flexibility makes it ideal for AI augmentation. A practical guide to overlaying GenAI capabilities on your existing ITIL framework.

Retrieval-Augmented Generation can turn your ServiceNow knowledge base into an intelligent self-service engine. Here's the architecture I recommend for enterprise ITSM knowledge automation.

Monitoring tells you when something you already anticipated breaks. Observability lets you ask questions you never thought to ask in advance. Most enterprise Splunk deployments are still doing the former.

Most organizations buy an AIOps platform before they know what maturity stage they're actually at. Here's the four-stage framework I use to diagnose where a team really stands before recommending anything.

Most enterprise GenAI-for-ITSM initiatives stall at a chatbot demo. Production-grade implementations look almost nothing like that — and the difference comes down to four specific design decisions.

A behind-the-scenes look at building Smart Desk — the architecture decisions, the failure modes I hit early, and why the boring parts of the build mattered more than the AI model choice.

Most ServiceNow instances accumulate governance debt the same way technical debt builds in codebases — quietly, through years of reasonable-seeming shortcuts, until an audit or an automation project hits the bill all at once.

Every API integration you add to ServiceNow expands the platform's attack surface, whether or not anyone treats it that way. Here's the checklist I run before any integration goes live.

A governance framework that only works during business hours, with the process owner available to clarify edge cases, isn't a governance framework — it's a documentation exercise waiting for its first real incident.

Change management gets a bad reputation because most implementations treat every change like it carries equal risk. ITIL 4's risk-based approach fixes this — if anyone actually implements it that way.

SLA governance at portfolio scale is a completely different discipline than SLA management for a single account. Here's what changes once you're responsible for fifteen accounts instead of one.

Most service transition failures aren't caused by the cutover itself — they're caused by hypercare planning that was treated as an afterthought rather than a core deliverable.