0%
Your team is getting new machines this cycle regardless — NPU laptops for most people, discrete-GPU workstations for the power users who run models locally. The only real question is whether they arrive with the AI silicon to run on-device copilots, or are obsolete the day your first one goes live.
On-device AI acceleration (the NPU) has moved from a premium add-on to standard silicon in a single hardware generation. Specify a machine without it now and you are buying a device that cannot run the local copilots your staff will expect within a year — at the same price as one that can.
A 500-person organisation on a three-year refresh cycle will spend roughly AED 1.0–1.6M on endpoints anyway — laptops for the field, workstations for finance, analytics and design. Specify non-AI machines and every unit becomes a device that cannot run local inference: staff keep round-tripping everything to the cloud, sensitive drafting leaves the building, and the power users who need local-model compute do not have it.
Do nothing and you commit to a second refresh inside 18 months to catch up — a full unplanned capex line — while the estate spends the entire gap without the productivity AI was supposed to deliver.
A representative planning scenario for a GCC mid-market organisation, built from the industry benchmarks cited below.
How to read this: the AI-silicon premium adds only a marginal amount to hardware you are already buying. The return is staff hours returned to the business plus the second refresh you avoid.
A deliberate sequence that proves value early and de-risks each step.
Map the estate by role, model the refresh you are already funding, and specify the right machine per persona.
Deploy a small AI-class fleet to representative users, with on-device copilots and local models enabled.
Roll out by department against the existing refresh schedule — no big-bang, no double spend.
Fold into managed lifecycle: imaging, security baseline, model updates and support on an agreed SLA.
Because we shape AI-class devices into the refresh you were already funding, there is no double spend and no new capex line. Central management and a security baseline are in from day one, and on-device processing keeps sensitive drafting and search off the cloud entirely.
The AI silicon is specified into the cycle you already budgeted — not a separate programme.
Rolled out department by department against the existing schedule, so cashflow is unchanged.
Sensitive drafting and retrieval stay on the machine — nothing leaves the building to be summarised.
Refresh cycles are landing now. Specify non-AI laptops and workstations today and you commit to buying again within 18 months — while your team spends that whole time without the productivity gain, and your power users wait even longer for local-model compute.
“Don’t buy laptops that are obsolete the day they land. Refresh once — into the AI era.”
No. We phase AI-class devices into your normal refresh as machines reach end of life, starting with the roles where the return is clearest. There is no need to write off working hardware.
For most staff the premium over the machine you were already buying is marginal, and the return is measured in hours per person per month plus the second refresh you avoid.
Both, specified by persona. NPU laptops cover the majority; discrete-GPU workstations go to the power users who run or tune models locally.
Yes — that is the point. Drafting, summarising and retrieval run on the machine, so sensitive content is not sent anywhere to be processed.
