01 / 06

CASE STUDY

Client Server

Rebuilding 25 years of precision recruitment workflows — without replacing the CRM
1998
Est. — 25+ Years in Niche IT Staffing
5 > 500
"Five perfect matches over five hundred fuzzy ones"
4 Phases
Quake Infrastructure Deployment
Quake · quake.ai Client Server · Case Study
Client Server 02 / 06

25 Years of Precision. One Migration That Broke Everything.

Client Server built a hyper-efficient niche IT recruitment operation inside a custom CRM over two decades. A mandated move to Mercury destroyed it overnight.

01

Who They Are

London's leading niche IT recruiter since 1998. Specialising in Java, Python, C++ and data science, their edge was precision: five perfect matches over five hundred fuzzy ones.

02

The Breaking Point

Mercury's one-size-fits-all architecture had no concept of validated skill taxonomies, single-screen interview capture, reverse matching, or time-based ownership rules. Every competitive advantage was gone.

Quake · quake.ai Client Server · Case Study
Client Server 03 / 06

Five Ways their CRM Broke Client Server

Every failure identified directly from discovery sessions with David Kerr, Operations Lead.

01

The Tag Soup Problem

Mercury collapsed RDB's structured skill taxonomy into an unordered list mixing locations, technologies and statuses. Precise searches became impossible — 5 perfect matches became 500 irrelevant ones.

02

The Lost Interview Plugin

RDB's single-screen plugin captured salary, location and validated skills in one flow during a live call. Mercury replaced this with 5+ separate tabs — unworkable mid-conversation.

03

No Reverse Search

Client Server's core workflow is candidate-first: get someone on the phone and instantly surface every matching live role. Mercury only supports job-first search — too slow for same-call conversion.

04

Broken Review List Logic

RDB had two distinct stages: Review List and Shortlist. Mercury collapsed them into one. A candidate could be reviewed six times for six roles — six consultants duplicating the same work.

05

The 2-Hour SLA Mercury Can't Enforce

When a job goes live, the candidate's owner has 2 hours to act — then any consultant can step in. Mercury has no timer, no escalation. Ownership disputes became a daily source of conflict.

Quake · quake.ai Client Server · Case Study
Client Server 04 / 06

The Quake Infrastructure Layer

Four targeted interventions deployed as a non-disruptive overlay on Mercury — each reversing a specific failure.

01

Two-Tier Validated Skill Architecture

AI parsing extracts broad skills automatically (Tier 1). A controlled validated list of core technologies forms Tier 2. Searches filter exclusively by verified tags — restoring the "five perfect matches" standard.

02

Interview Plugin 2.0

A bespoke single-screen interface built on Quake Core. Pulls the candidate from Mercury, captures all fields in one flow during the call, pushes back via API — Mercury's native UI bypassed entirely.

03

Bidirectional Vector Search

While the consultant is live on a call, Quake runs a background vector search against all open jobs and presents ranked matches in real time. Candidate-first matching fully restored — now AI-powered.

04

Review List Logic & SLA Engine

Candidates rejected against a role are hidden from future queues — eliminating duplicated work. The 2-hour ownership timer is enforced at infrastructure level: windows close automatically.

Quake · quake.ai Client Server · Case Study
Client Server 05 / 06

From Friction to Flow

Non-disruptive overlay on Mercury. Results visible from the first production shortlist.

Client Server
3.4×
More qualified candidates per brief

Validated tagging restored precision — surfacing the right Java engineers, not adjacent JavaScript developers.

Client Server
62%
Faster time-to-shortlist

Vector search running live during candidate calls compresses days of matching into under an hour.

Client Server
99%
CV extraction accuracy

Hard-wall sandbox replaced three failing integrations with a single, auditable parsing pipeline.

Client Server
Zero
Ownership disputes since SLA launch

2-hour timer enforced at infrastructure layer — candidates unlock automatically, conflicts eliminated.

Quake · quake.ai Client Server · Case Study
Client Server 06 / 06

The Foundation for an Agentic Business

Precision workflows restored — and now the foundation for the next stage: AI agents that do the legwork without breaking the ownership model.

01

David Kerr, Operations Lead

"The ownership stays stable. AI agents come along, do the legwork, tap the consultant on the shoulder: I've got three candidates — why don't you send these out? The candidate gets a great experience because the same person owns that relationship throughout."

02

What This Unlocks

With stable ownership enforced at infrastructure level, Quake's agentic workflows operate safely. The AI-first autopilot — AI matches, consultant edits — becomes viable without litigation risk. Client Server is positioned as one of the UK's first fully agentic recruitment desks.

Quake · quake.ai Client Server · Case Study