IT & Networking.
The nervous system that everything else needs.

Networking is the most important layer in a smart building because every CCTV camera, access panel, BMS controller, automation hub and PBX runs on it — a weak network surfaces as failures everywhere. Every other system runs on it. CCTV cameras, access control panels, BMS controllers, automation hubs, telephony, video conferencing, building Wi-Fi — all need a stable network with adequate bandwidth, PoE capacity, and redundancy. A weak network surfaces as failures in every other system, but is rarely diagnosed there.

Cable a new building with Cat6A — the premium over Cat6 is under 10% and it supports 10 Gbps over a full 100 m channel for 25 years. The premium over Cat6 is now small — under 10% — and it supports 10 Gbps to the desk over the full 100 m channel without compromise. It is rated for 25 years of service. Cat6 should be reserved for retrofit projects where pathway constraints force the choice.

Yes — Wi-Fi 7 is worth paying for on any new deployment in 2026; on recent 6E installs, wait one refresh cycle. For any new deployment, yes. The access-point premium over Wi-Fi 6E is modest and shrinking each quarter. The 6 GHz band, MLO multi-link operation and 320 MHz channels deliver real-world capacity gains, particularly in dense rooms (boardrooms, classrooms, lobbies). For retrofit on a recent 6E install, wait one refresh cycle.

Approach SMB network security in three layers — VLAN segmentation that contains breaches, EDR/XDR on every important endpoint, and immutable offline backups with tested recovery. Segmentation that contains breach lateral movement (separate VLANs for IoT, building systems, guest, staff). EDR or XDR on every endpoint that matters. Immutable backups stored offline with documented recovery testing. The firewall is necessary but not sufficient — design assumes the bad day will arrive, and the question is recovery speed.

Hybrid for almost every organisation in 2026 — pure cloud only suits very small teams or consumer-internet companies; the middle stays hybrid. Almost every organisation in 2026. Latency-sensitive and data-residency-bound workloads stay on-prem; analytical and bursty workloads move to cloud. Pure-cloud strategies make sense at very small scale (under 50 employees) and very specific large scale (consumer-internet companies). The middle is hybrid, and the architecture rewards good network design.

Wi-Fi is sized through a predictive RF survey, validated on-site once the structural shell is up, modelled for year-three concurrent device count and targeted at -65 dBm at every desk. A predictive RF survey on the architect's floor plan, validated by an on-site walk-through once the structural shell is up. We model concurrent device count for the year-three load (laptops, phones, IoT, AV endpoints, building systems), target a minimum -65 dBm at every desk and -70 dBm in transit zones, and place access points so adjacent cells overlap by 20–25% for seamless roaming. The deliverable is a heat-map drawing the IT team can actually verify post-install.

Build network redundancy at the layers where downtime hurts: dual ISP links, stacked core switches, dual-homed servers and online double-conversion UPS — to a stated recovery target. At the layers where downtime hurts. Dual ISP links from physically diverse paths with automated failover. Stacked or paired core switches. Dual-homed servers and storage. Run-to-failure UPS architecture is replaced by online double-conversion with battery autonomy. The aim is not zero downtime — that is uneconomic — but a stated recovery target that the design honours, with the failure modes documented for the operations team.

Every patch, port and outlet is labelled to a published room-rack-panel-port schedule, with rack elevations, fibre splice records, OTDR traces and TIA channel-test certificates in the as-built pack. Labelled to a published schedule — typically room-rack-panel-port format, agreed with the IT team before deployment. The as-built pack contains rack elevations, patch schedules, fibre splice records, OTDR test traces and channel-test certificates per the relevant TIA standard. Years from now, when someone needs to find which port serves which desk, the answer is in the documentation rather than in someone's memory.