I don't evaluate technology from vendor slide decks. I build with it, test it in real conditions, and make decisions based on what actually works at scale.
Every technology decision runs through the same framework, whether it's a connectivity overhaul across a global fleet or a new AI tool for a single workflow. The questions don't change. The scale does.
Not a theoretical one. Not a vendor-manufactured one. I start with what is actually breaking, what is costing too much, or what is slowing people down. No clear operational pain, no project.
Purchase price is the smallest number. I evaluate the total cost: licensing, training, support burden, integration complexity, and what happens when something fails at 2 AM on a vessel with no on-site IT.
A fix for one location is a workaround. A standard that deploys across an entire fleet is infrastructure. I test in constrained environments first. If it cannot survive limited bandwidth and remote support, it does not ship.
The best technology fails if the team cannot maintain it. Supportability, documentation requirements, and training investment are part of every evaluation. Every solution has to work without any single person in the room.
Same questions every time. The judgment is knowing when the answers are good enough to move forward.
The projects I take on share a common theme: find the inefficiency, build a standard, deliver improvements with little to no net new cost.
A procurement platform that automates manual handoffs while giving every stakeholder the view they need: technicians see receiving status, FinOps tracks spending, and the CISO and PMO get a global dashboard. I built the first version solo across a week of evenings. AI-assisted development closed the last twenty percent in under an hour. Fewer people touch the process; everyone sees the full picture.
The question for every AI tool: does this free the team to work on things that affect the business, or is it just a new thing to manage? Engineers should solve problems that affect operations and revenue, not reset passwords and chase SSO tickets. AI earns its place when it eliminates the menial work.
Not a project with an end date. An ongoing discipline. Swapping patch cables for shorter runs so racks are cleaner. Reallocating equipment between sites instead of buying new. The best improvements are low effort, low cost, and compound over time without requiring budget approval.