How Braintrust Uses Claude to Speed Hiring and Cut Screening Costs

Recruiting platforms face a fundamental volume problem: the roles that attract the most applications are precisely the ones where manual screening breaks down fastest. Braintrust, a recruiting technology company, was building toward a future where AI could handle the repeatable parts of hiring—screening, initial interviews, matching—while freeing recruiters to focus on judgment-intensive work.

The challenge was acute in both directions. Employers were drowning in low-quality applications—as many as 800 candidates for a single role, with only 10% credible matches. Job seekers, meanwhile, were submitting applications to roles they had little visibility into, often filtered out before any human interaction. The system was inefficient for both parties, and the company needed AI that could handle professional conversations at scale without introducing bias or producing unreliable output.

Braintrust evaluated multiple large language models before selecting Claude. The decision came down to three factors: Claude’s commitment to ethical, unbiased output for professional tasks like interviews and job descriptions; a 200,000-token context window large enough to support 30- to 60-minute interview transcripts; and native compatibility with Amazon Bedrock, which matched the company’s existing AWS infrastructure. Deployment leveraged AWS CloudWatch for monitoring from the start.

Three core features were built on Claude. AIR (AI Recruiter), powered by Claude 3.5 Sonnet, conducts initial screening interviews autonomously—giving every applicant a real conversation and providing recruiters a structured assessment. A Job Description Generator creates richer, more targeted postings that attract qualified candidates. A semantic search tool in development allows natural-language queries across the candidate database rather than rigid filter sets. The combined result: 100% of initial talent screens are AI-powered, saving more than $150K to date, and job postings using the AI generator attract 25% more applicants. Nearly 90% of actual hires come from candidates rated as good or great matches by the system.

Braintrust’s leadership sees this as the beginning of a structural shift in how hiring works. With AI handling the screening layer, recruiters can focus on relationships and storytelling—explaining the company’s culture and values to candidates who have already been validated. The company also sees Claude as a path toward more equitable hiring, since LLM-based evaluation is less subject to the pattern-matching biases that accumulate in human reviewers over time.