01
/ 08
CASE STUDY
Leaders in Care & Fintelligent
A headless Quake transformation — from high-risk "Shadow AI" to a governed, agentic ecosystem across the Bullhorn, SourceFlow and Idibu stack
Challenge
02
/ 08
The Engagement
Leaders in Care (healthcare) and Fintelligent (financial services) were moving fast on ungoverned "Shadow AI" — consultants pasting candidate data into personal ChatGPT tabs. The mandate: keep the speed, but institutionalise the data inside the CRM.
The Challenge
Fragmented DIY processes and brittle keyword search across a 150,000+ record Bullhorn estate. Nurses applying at 3 AM were placed by competitors before the 9 AM office start, and GDPR-sensitive data was leaking into unmanaged LLM tools outside the CRM.
How Quake Helped
Quake Core was deployed as headless "plumbing" directly onto the existing Bullhorn, SourceFlow and Idibu stack — a centralised, API-led infrastructure with vector search and a ring-fenced AI portal that unlocked agentic workflows with zero system migration.
Infrastructure
03
/ 08
Core Infrastructure & Data Environment
Months 1–2 — establishing the Quake Core "plumbing": a centralised, API-led foundation replacing fragmented DIY processes.
Bullhorn Integration & Ingestion
A seamless API connection to Bullhorn imported ~150,000–160,000 candidate records into a single governed data layer.
Multi-Instance Segregation
Separate Quake instances for Leaders in Care and Fintelligent preserve data integrity across both brands, with a centralised "Login with Microsoft" protocol enforcing security and blocking unauthorised Gmail logins.
Vector Search & Context Mapping
Semantic vector search reads clinical nuance and intent — matching a "Complex Care Nurse" to a specific shift requirement instead of relying on brittle keyword search.
Historical Data Synchronisation
Job sync dates standardised from Jan 01 2024 for reporting accuracy, while older candidate and client records were pulled to give the system full database context.
Quake Core
04
/ 08
Headless Infrastructure
Quake Core: The Plumbing Behind Both Brands
A centralised, API-led data layer wrapped straight around Bullhorn, SourceFlow and Idibu — replacing fragmented DIY processes with one governed engine for Leaders in Care and Fintelligent.
Wrap Bullhorn, don't migrate
Quake pointed at the existing Bullhorn estate and ingested ~160,000 candidate records — agents ran on it within days, with no rip-and-replace and Bullhorn left as the system of record.
Two brands, segregated by design
Separate Quake instances keep Leaders in Care and Fintelligent data fully isolated, with centralised "Login with Microsoft" enforcing access and blocking unauthorised Gmail logins.
Vector search built for AI workers
Sub-second semantic reads let agents grasp clinical intent — matching a "Complex Care Nurse" to a specific shift — instead of breaking on brittle keyword search.
Agentic Workflows
05
/ 08
Deployed Agentic Workflows
Months 3–4 — custom agents built to clear the specific operational bottlenecks surfaced during the consultancy phase.
The "3 AM Screener" — Voice Interceptor
An unscripted AI voice agent (Vapi/Twilio) intercepts Idibu/SourceFlow applications and calls the candidate within 5 minutes — screening for clinical criteria such as PIN number and shift availability, then booking applicants straight into consultant diaries before competitors reach them.
The "Matchmaker" & Outreach Engine
Automatically matches incoming CVs against live jobs scraped from global boards or Bullhorn, then drafts customised speculative emails to hiring managers — saved as drafts in Office 365 for consultant approval.
Shadow AI Defence — Branded Portal
A secure, ring-fenced AI portal keeps the speed of LLMs for job-spec writing and CV formatting, while all data stays GDPR-compliant and institutionalised in the CRM rather than trapped in personal ChatGPT tabs.
Platform
06
/ 08
One Governed API. The Entire Recruitment Stack.
Every record and every AI agent is addressable through a single, governed API — the control plane enterprises deploy AI into.
One surface, not a hundred integrations
Build once against Quake instead of stitching together brittle point-to-point connections to every tool in the stack.
Governed for enterprise
OAuth 2.0, 75 granular permission scopes and tenant isolation — the controls regulated buyers require before deploying AI.
The whole recruitment model
Candidates, jobs, placements, agents and more — the entire stack runs through one addressable graph.
Efficiency
07
/ 08
Operational Efficiency & Administrative Agents
Months 5–6 — removing manual "toil" and retiring high legacy licence costs.
M3R "Data Cleaner" (Oracle Agent)
Automates the daily 6 AM download, reformat and rename of BI spreadsheets at 100% accuracy — eliminating the manual data-entry errors that previously crashed the system.
Sense & HireVue Replacement
Linear automation journeys such as "Silver Medalist" re-engagement replicated into Quake's agentic framework, positioning the firm to decommission these legacy tools by Q4 2026.
Post-Call Interface — The "30-Second Rule"
Ringover call transcripts are ingested and surfaced as a human-in-the-loop pop-up: the AI suggests tags, the consultant clicks "approve", and Bullhorn updates instantly — fixing chronic CRM data hygiene.
Roadmap
08
/ 08
Projected Impact & Strategic Roadmap
Technical deployment is ahead of schedule. Per Taya Hornsby, the priority remains 100% data accuracy before full workflow scaling — to avoid "useless journeys".
Q1–Q2 · Revenue & Foundation
100% data governance established and instant response to 3 AM applicants — closing the out-of-hours placement gap.
Q3 · Monetisation
Database reactivation and a measurable CRM data-quality lift driven by Ringover transcript ingestion.
Q4 · Structural Shift
Full BI replacement and automated commission reporting, decommissioning legacy Sense and HireVue tooling.