← All projects Case Study · VanDercook Ventures

Ellie, an AI hospitality recruiter that starts with a conversation

She places one phone call, runs a structured eight-minute interview, then makes the warm introduction. No forms, no resume uploads, no cold recruiter commission. The candidate pays nothing; the operator pays only when a hire lands.

Role  Founder · Full build Category  AI voice agent · Two-sided marketplace Stage  Live beta meet-ellie.com ↗

The problem

Hospitality runs on people, and hospitality management hiring is broken at the speed of email. Operators get buried in resumes that don't fit the room. GMs and chefs scroll Indeed listings written for the wrong city, wrong pace, wrong concept. Recruiters charge fifteen to twenty percent of first-year salary for cold introductions that never set foot in the venue. Job boards capture form data and leave both sides guessing.

The category needs a recruiter who lives in the trade, listens to who you actually are, and only surfaces matches that are real. Ellie is that recruiter, built as software.

The product

A hospitality manager or operator visits meet-ellie.com, picks a side ("I'm looking" or "I'm hiring"), and submits their phone number. About a minute later, Ellie calls. The call is eight to ten minutes. She introduces herself, discloses she's an AI agent, then asks one open-ended question at a time and listens, following a locked interview ladder for each side.

When she finds a real match, she sends both sides a warm introduction email. No portal full of unread roles. No recruiter phoning in a commission. The whole flow is built on a single principle:

The call is the means, the profile is the product. What Ellie hears in eight minutes is signal that no job board ever captures.

How it works

1

Submit

A phone form on either side of the marketplace. SMS opt-in is separate and optional. The form delays the outbound call 30 to 60 seconds to give the person a chance to defer.

2

Talk

Ellie calls, opens warm, discloses she's an AI agent, and asks questions from a structured tree, but the conversation is generative. She follows up on what's actually said, never list-asks, keeps her turns short. Voicemail? She hangs up cleanly and sends a recovery text the same day.

3

Match

The call is transcribed, summarized, and structured into the database. Each candidate-to-posting match is scored with a written rationale and confidence score. An operator reviews matches in a console before any introduction, Ellie never makes a cold intro.

4

Introduce

On approval, a warm two-sided introduction email goes to both parties at once. Five days later, Ellie checks in: one-tap reply, met / not yet / won't happen. Five days after that, a follow-up if no meeting has landed.

5

Retain

After a placement, Ellie checks in at day 30, 60, 90, and six months, one-tap reply on how the home is settling for both sides. The celebration email asks the new hire for a single referral. The grapevine has always been hospitality's best hiring channel.

What makes Ellie different

A voice agent built for the industry

She uses industry vocabulary natively, asks about your "room," your "pass," your "house," knows a sous chef from a chef de cuisine. The ladder captures the leadership story resumes never carry.

Honest pricing

A flat $1,000 placement fee, or a $2,500 monthly retainer for three-role coverage. Operators pay only when a hire lands. A contingency recruiter at a $90k role costs roughly $16,200. The math is part of the pitch.

A single phone call

No forms. No resume uploads required. No job board to scroll. The candidate's whole interaction before any introduction is one conversation that fits in a coffee break.

Profile provenance

Every field is tagged with how it got there, the call, the candidate's edits, or LinkedIn enrichment via Apollo. Data quality is treated as a partnership, not a one-way capture.

An editorial product, not a SaaS spreadsheet. Pricing reads as a tasting menu (Counter, Bar, Chef's Table). The operator dashboard reads as "tonight's pass." Matches surface like a chef's resume with a "why this match" footnote.

Architecture

VoiceTwilio ConversationRelay with Async Answering Machine Detection
Voice agentClaude with a locked system prompt and a structured interview tree
BackendNode.js / Express / TypeScript on Railway
DatabasePostgres
EmailResend, with delivery webhooks for bounce + complaint tracking
SMSTwilio Programmable Messaging, 10DLC compliant
FrontendReact + Vite + TailwindCSS
Data enrichmentApollo.io People Match (on LinkedIn URL)
AuthenticationHMAC magic-link tokens for portals; operator-passcode header for the console

The entire codebase passes a custom em-dash lint gate, a build-fail rule that enforces brand-voice consistency in every customer-facing string.

What's been shipped

Roadmap

Visit Ellie Next project: A³ →