AI in IT has two modes: genuinely useful and massively overhyped. Here's where it lands in my actual workflow.
Ollama runs large language models on your own hardware. I use it for:
The models I keep loaded: Llama 3.1 8B for fast queries, Qwen2.5-Coder 14B for code tasks, Mistral 7B when I want a second opinion.
Setup takes ten minutes on any machine with 16GB RAM. A dedicated GPU speeds up inference significantly — the NVIDIA RTX 3060 12GB handles 13B models at usable speed.
When I need multi-step reasoning — designing a network architecture, reviewing a firewall ruleset for gaps, debugging a complex Ansible playbook — I use Claude or GPT-4o. Local models at 7–13B can't match the reasoning depth here yet.
Fabric is a command-line tool that pipes text through AI prompts. My daily use: pipe a log file through a "summarize errors" prompt, or pipe a long config into a "spot the security issues" pattern.
cat /var/log/syslog | fabric --pattern analyze_logsIt connects to Ollama or any OpenAI-compatible API.
n8n is a self-hosted automation tool with an AI agent node. I use it for:
Run it as a Docker container. The n8n + Ollama combo on a mini PC handles all of this offline.
Asking AI to diagnose production outages in real time. It hallucinates plausible-sounding but wrong command flags. For live troubleshooting, documentation and man pages are still faster. AI is a draft tool, not an oracle.
The Kingston Fury 32GB DDR4 kit is the upgrade I'd make first if you're RAM-limited.
As an Amazon Associate, IT Lad earns from qualifying purchases made through links on this page. Read the full affiliate disclosure.