Episode 170
AI in Recruitment: Automating with Impact at ThoughtWorks
Marcus Thorpe, Global Head of TA at ThoughtWorks, reveals how agentic AI transforms recruiting workflows—not by replacing recruiters, but by freeing them to build genuine candidate relationships. Learn which automation bets actually move the needle on time-to-hire and offer acceptance.
Episode Key Takeaways
Quasi-empathy isn’t empathy. Even though AI can appear empathetic, it lacks genuine understanding of why candidates leave jobs or choose one role over another. The best recruiters still win by building real relationships—AI handles the administrative friction so they can focus on that human connection.
Interview feedback quality was the hidden blocker. Before BrightHire, ThoughtWorks had thumbs-up/thumbs-down feedback with almost no substance for unsuccessful candidates. AI transcription and auto-population of attribute-based feedback eliminated 15 minutes of post-interview admin per interviewer, freeing capacity to maximize every interview.
Scheduling software solved a fairness problem, not just a time problem. Manual coordinator matching meant some interviewers gamed the system by removing skills from their profiles to avoid certain languages. Automated matching with clear parameters—like ‘no more than one interview per week per person’—distributed load fairly and improved interviewer availability.
Dynamic Boolean searches arm every recruiter with sourcing COE expertise. Instead of running the same LinkedIn search repeatedly, agentic AI refines Boolean strings based on success/failure patterns, response rates, and company-specific factors. The goal: scale sourcing capability across the entire team, not replace the sourcing team.
Time-to-offer-accept directly drives revenue. At a professional services firm where consultants are billed to clients, every day saved in hiring is a day of extra billing. Reducing the face-to-face interview stage—which consumed half of total time-to-hire—translates to measurable commercial impact.
Frequently
Asked
Questions
What percentage of candidates opt out of AI-assisted interview recording?
Globally, opt-out rates are 2–3%, well below the response rate threshold. Rollouts in IME and LATAM showed strong acceptance. Germany is expected to be more resistant, though the feature hasn’t launched there yet. Candidates appreciate the conversational experience when interviewers aren’t taking notes constantly.
How do you justify automation spend to the CFO without headcount reduction?
The business case centers on capacity and speed, not cost-cutting. Faster ramp for new coordinators (weeks vs. months), reduced time-to-offer-accept, and ability to handle 4,000+ hires annually without proportional team growth. At ThoughtWorks, every day saved in hiring translates to additional client billing revenue.
What's the risk of over-automating the candidate experience?
Over-automation removes the personal touch that closes offers. Best candidates interview simultaneously at multiple companies. Recruiters must automate back-office tasks—scheduling, feedback capture, offer generation—while preserving high-touch relationship-building. The balance is critical: automate what doesn’t touch candidates, humanize what does.
How does AI improve interview training at scale?
GoodTime integrates with the LMS to track completion and automatically trigger shadowing journeys. With a 4–5:1 pass-through rate, traditional shadowing often misses positive examples. Virtual shadows using Notebook LLM create anonymized, reusable interview scenarios that can be updated instantly without re-filming, scaling best-practice training across thousands of interviewers.
What's the sourcing bottleneck when you're hiring 4,000+ people annually?
Sourcing COEs in India are finite in throughput. Recruiters vary in sourcing skill. Agentic AI refines Boolean searches based on success factors, response rates, and client-specific parameters, allowing all recruiters to access the same expert-level capability. Goal: 70% of hires from sourced or referral channels, reducing reliance on inbound applications.