D-02 · UK technology group case study
A UK technology group with a holding company, a public-policy advisory firm and an AI intelligence platform needed all three live, consistent and fast.
The short answer
AiS LABZ designed and built the digital estate for a UK technology group working in government, policy and AI: a group site, an advisory firm site and an AI platform site, each on its own design system. After a performance rebuild the platform site's mobile score rose from 44–56 to 87 and first paint fell from 9.2 to 1.6 seconds.
What was in the way?
- Three brands with no shared standards for type, colour or components.
- A 2.2 MB single-file build painting nothing on mobile for over nine seconds.
What changed?
- Mobile score 44–56 to 87 on the platform site; desktop 61 to 87–98.
- First paint 9.2s to 1.6s on mobile.
- Platform HTML from 2.2 MB to 140 KB, 30 KB over the wire.
- Advisory site HTML down 42%, mobile score 81 to 87.
- Enquiry delivery verified end to end, not just accepted by the server.
How did we do it?
The same four-phase method on every engagement: diagnose, model, build, compound.
Diagnose
Each brand kit, audience and job, with senior stakeholders at group level.
Model
One estate, three identities: shared engineering standards, distinct design systems.
Build
Three design systems of 14–18 components and 88–110 tokens, three responsive sites, enquiry and demonstration forms routed to the group inbox, and GA4 and Search Console on each.
Compound
A performance rebuild: platform HTML from 2.2 MB to 140 KB, and self-hosted fonts across the estate.
Who did we work with?
What happens next?
Insight and article programmes on real URL paths, so each page can be indexed on its own.
Common questions
Client confidentiality. The work, the numbers and the seniority of the people involved are real.
Each brand serves a different buyer. Shared engineering keeps them fast; separate systems keep them distinct.
Removing embedded assets from the HTML, self-hosting fonts and deferring non-critical visuals.
Your number next
