Information systems · AI delivery · self-hosted
Nearly twenty years in enterprise IT and retail operations, most of it spent making systems talk to each other. Now that means AI: a business intelligence platform that reads the messy half of the business, an organisation-wide Copilot rollout, and a self-hosted agent in my lounge room doing the same job at a scale I'm allowed to break.
For most of my career, if information wasn't sitting neatly in a database column, the system couldn't do much with it. The reasoning lived in emails, the exception lived in someone's notes, and the real answer lived in a document nobody had opened since 2019. So the business ran on clean numbers plus separate opinions about what they meant.
AI changes that. Software can finally read the messy half — notes, emails, the edge cases people quietly work around — and turn it into something structured enough to act on.
I don't want a chatbot to keep me company. I just dislike doing the same administrative triage twice.
Give a small model a narrow brief, a strict set of tools and somewhere to write its notes, and it quietly handles the chores I used to do by hand. Not intelligence — leverage.
I lead the information systems and business improvement function in a national organisation — roadmap, vendor engagement, IT budgeting and policy, and the governance underneath it. Half the job is delivery; the other half is explaining what it will cost, what it won't fix, and what has to change first.
Most organisations keep clean numbers in SQL tables and the actual reasons buried in documents. I designed and built a business intelligence platform that uses AI to read both at once — structured metrics and unstructured operational text in a single reporting view, replacing a handful of fragmented sources.
Organisation-wide Microsoft Copilot from rollout through adoption, including specialised agents designed and deployed for specific business functions on top of a governed process library.
AI doesn't fix undocumented processes or messy permissions — it finds them faster. Underneath sits the actual IT: a full laptop and mobile fleet migrated into Intune, Entra ID and Microsoft 365 administration, SharePoint and automated flows, Essential Eight control uplift.
One mini-PC with no GPU, a house full of sensors, and an agent that never sleeps. Same problems as work, smaller blast radius.
Runs continuously on the host: watches the homelab, keeps a written commitment ledger, drafts the morning brief, and messages my phone when something needs a human. Loads 163 skills on demand, and every fix becomes a note so the next attempt starts where the last one finished.
247 live entities across 16 integrations — solar, battery, grid, the car and the charger in one view. The house now makes electricity decisions I used to make by guessing.
A small model runs on the box next to everything else. No API bill, no data leaving the house. Slower than the big ones, and that's the point — it's how I find out where the technology stalls before I put anyone's workflow near it.
Not selling consulting, no newsletter. If you're running enterprise systems, building practical agents, or trying to get useful work out of small hardware, I'm happy to compare notes. LinkedIn keeps the spam out and lets me see who's knocking.