Build a Voice AI That Prequalifies Recruiting Candidates
Syntora is a consultancy that builds custom voice AI for candidate prequalification. We deploy systems that screen applicants over the phone using conversational AI.
Scope depends on the number of screening questions and the specific Applicant Tracking System (ATS) you use. A 5-question screener that posts a summary to Greenhouse is a 2-week build. A 10-question screener with conditional logic that integrates with a custom API requires more discovery and a 4-week build cycle.
We built a voice prequalification system for a 12-person recruiting firm processing 400 applicants a month. Recruiters were spending 15 minutes per candidate on initial phone screens. The AI went live in 2 weeks, handling the 5 core qualification questions and reducing recruiter screening time by 80%.
What Problem Does This Solve?
Most recruiting teams try to automate screening with basic tools that fail in practice. Standard IVR (Interactive Voice Response) systems are the first attempt. They use rigid "press 1 for yes" menus that cannot handle natural conversation. A candidate who says "I have about five years of experience" instead of just "five" breaks the workflow, leading to a 60% drop-off rate and useless data.
A second approach is using a scheduling tool with screening questions in the form. This just pushes the manual work around. Instead of screening on the phone, the recruiter now sifts through dozens of form submissions filled with unqualified candidates who can book calendar slots. This clogs the pipeline and wastes interview time with people who don't meet basic requirements, like having a valid driver's license for a field position.
These tools fail because they are not conversational. They are glorified forms. They cannot ask clarifying questions, understand context, or interpret nuanced answers. Real qualification requires a system that can have a basic, stateful conversation, which is an engineering problem, not a configuration problem.
How Does It Work?
First, we map your exact prequalification script to a conversational flow graph. We identify the 5-7 critical data points you need to collect and define how they map to fields in your ATS, whether it is Greenhouse, Lever, or a custom-built system. We use the Claude API to script conversational paths that feel natural and can handle unexpected answers from candidates.
The core system is a Python application built with FastAPI. It uses a Supabase Postgres database to maintain conversational state for each candidate, allowing it to reference previous answers in follow-up questions. For telephony, we integrate directly with Twilio's API to place and manage the outbound calls. The AI's responses are generated by the Claude API and returned in under 800ms to keep the conversation fluid.
The entire service is deployed as a serverless function on AWS Lambda. When a new candidate applies through your ATS, a webhook triggers the Lambda function, which initiates the prequalification call. After the 3-5 minute call, the service generates a structured summary and a full transcript, then uses the ATS API to post them directly to the candidate's profile. This entire process, from call completion to ATS update, takes less than 15 seconds.
We implement structured logging with `structlog`, sending JSON logs to AWS CloudWatch. This allows us to set up specific alerts. For example, if the ATS API integration fails more than 3 times in an hour or the call completion rate drops below 90%, a notification is sent to a designated Slack channel for immediate review. You get full visibility into the system's performance.
What Are the Key Benefits?
Get Candidate Transcripts in 2 Weeks
Go from your current manual process to receiving the first AI-generated screening transcripts in your ATS in 10 business days.
Fixed-Price Build, Not Per-Call Fees
One fixed project cost and minimal monthly hosting under $50 for hundreds of calls. No unpredictable SaaS bills that scale with applicant volume.
You Own All the Source Code
We deliver the complete Python codebase to your company's GitHub repository. You are free to modify or extend it without vendor lock-in.
Get Failure Alerts Directly in Slack
We configure monitoring to alert your team via Slack if the ATS integration fails or the AI struggles to understand candidate responses.
Writes Directly to Your Existing ATS
The system integrates with Greenhouse, Lever, or any ATS with an API. Recruiters see transcripts and summaries in the tool they already use.
What Does the Process Look Like?
Script & ATS Access (Week 1)
You provide your screening questions and grant read/write API access to your ATS. We deliver a technical specification and data mapping document for your approval.
Core AI Build & Demo (Week 2)
We build the conversational agent and integration logic. You receive a private phone number to call and test the screening experience yourself.
Deployment & Live Testing (Week 3)
We deploy the system on your cloud infrastructure and connect it to your live ATS. The first 20 real candidates are processed with our team actively monitoring each call.
Monitoring & Handoff (Weeks 4-6)
We monitor 100% of calls, tune the AI for edge cases, and finalize documentation. You receive the full source code and a runbook covering maintenance and updates.
Frequently Asked Questions
- How much does a custom voice AI cost and how long does it take?
- The build timeline is typically 2-4 weeks. Cost depends on the number of screening questions, the complexity of conversational logic, and the specific ATS integration. A simple 5-question screener for Greenhouse is a smaller scope than a 10-question screener with branching logic for a custom-built system. We provide a fixed-price quote after a discovery call.
- What happens if the AI misunderstands a candidate or the call drops?
- If the AI has low confidence in its interpretation of an answer, it flags the transcript for human review in the ATS. If a call drops, the system can be configured to retry once after 10 minutes. All partial transcripts and call logs are saved, so a recruiter always sees what happened and can follow up manually.
- How is this different from a platform like Talkpush?
- Talkpush is a SaaS platform where you are locked into their features and per-user pricing. We build a dedicated voice system that you own completely. It runs on your infrastructure, integrates deeply with your specific workflow, and has no per-seat or per-candidate fees. You get the full source code and control the entire process.
- Will this sound robotic and turn candidates off?
- We use modern text-to-speech models and the Claude API for natural language generation. The agent is conversational, not a rigid IVR. The goal is not to trick candidates, but to provide a fast, efficient screening they can complete 24/7. Most candidates appreciate the ability to get through the initial step without scheduling a call.
- What happens when the AI models you use are updated?
- Our code uses standard API calls to providers like Anthropic. When they release a better model, updating is typically changing a single line in a configuration file (e.g., from 'claude-3-sonnet' to 'claude-3.5-sonnet'). We document this simple process in the runbook, allowing you to upgrade to the latest technology without a rebuild.
- How is sensitive candidate personal information handled?
- The system is deployed on your own cloud infrastructure, not Syntora's. Candidate data is processed in-memory and passed directly between the telephony API, the AI model, and your ATS. We do not store any Personally Identifiable Information (PII) on our systems. Audio recordings can be set to auto-delete after transcription to meet compliance requirements.
Related Solutions
Ready to Automate Your Small Business Operations?
Book a call to discuss how we can implement ai automation for your small business business.
Book a Call