HomeExecutive overview
Under the hood · how it was built

A prototype that catches candidates going cold.

An AI prototype that detects candidates going cold at any stage of the SmartRecruiters hiring funnel and proactively re-engages them on WhatsApp — with recruiters always in control. Here is how it was built, how it runs, and where it goes next.

The runtime loop
Detect1cold at any stageReach2WhatsApp templateConverse3adaptive AI chatRecover4or hand to recruiter
How it was built

Five gates. Each reviewed and approved before the next began.

This prototype was delivered through an approval-driven, phase-gated process. Select a gate to see what it covered.

01

Scope

Gate 1 of 5 · The gap + the boundaries
Reviewed & approved

Name the problem precisely and draw the lines around it.

  • Pinpointed the gap: hiring automation only moves candidates forward — nothing brings back the ones who go quiet.
  • Set boundaries up front: the agent re-engages, but never advances, rejects, or hires. Every stage move stays a recruiter action.
  • Defined the proof-of-concept frame — a self-contained demo with zero Sutherland data and simulated integrations.
The prototype workflow

From cold signal to recovered candidate — in four moves.

Detect → Reach → Converse → Recover. A continuous loop that brings quiet candidates back without ever taking a hiring decision out of human hands.

01

Detect

A 5-minute sweep reads time-in-stage across the whole funnel and flags candidates going cold — computed in the prototype, since the ATS emits no native "cold" event.

02

Reach

A pre-approved, compliant WhatsApp template re-opens the conversation. The candidate’s reply opens a 24-hour window for free-form, adaptive chat.

03

Converse

Winston, the AI assistant, runs a warm, personalized, multi-turn conversation — gathering answers or booking time via in-chat WhatsApp Flows.

04

Recover

It logs the outcome back to the ATS and alerts the recruiter. It never auto-advances or rejects — every stage move stays human.

Recruiter-controlledThe agent re-engages on its own, but has no tool to move a candidate’s stage. Advancing, rejecting, or hiring is always a recruiter action.
Envision the finished product

What this becomes beyond the prototype.

The items below are the production roadmap — a forward look, not current POC features. The prototype proves the workflow today; production turns the simulated surfaces into live, always-on ones.

Roadmap

Live integrations

Swap the simulated ATS and WhatsApp surfaces for live SmartRecruiters and WhatsApp Business connections — a binding swap behind the same workflow.

Roadmap

Autonomous scheduling

Move the detection sweep onto always-on cloud crons so re-engagement runs continuously, with no one pressing a button.

Roadmap

Multi-channel reach

Extend beyond WhatsApp to SMS, email, and web chat, meeting each candidate on the channel they actually answer.

Roadmap

Deeper analytics

Layer in recovery-rate, time-to-reconnect, and cohort reporting so the business can see exactly what re-engagement returns.

Keep exploring

See how it’s grounded — then watch it work.

The foundations page covers the public-domain research it was grounded in and the cloud infrastructure it runs on, with zero Sutherland data. The dashboard shows the live prototype in action.

How it’s grounded & hosted →Open the live dashboardHome