Mishcon de Reya brings an AI chatbot into first-round interviews
A high-profile London law firm is introducing an AI chatbot to run first-round interviews for graduate applicants. Partners at Mishcon de Reya call it a landmark step for a legal practice, aiming to handle large applicant volumes while giving each candidate a more personal conversation.
The chatbot uses details from a candidate's application to initiate an interview and generate a transcript. The firm's early careers team then reviews each transcript before moving candidates forward.
How the interview works
The AI pulls information from submitted applications, asks follow-up questions, and keeps the conversation moving. Graduates get more room to share their story than a fixed character limit allows, and recruiters get a structured record to compare across applicants.
Early careers manager Tom Wicksteed told The Times the format helps candidates expand on achievements, experiences and motivations, while giving recruiters better signal earlier in the process.
Who built it
The tool comes from graduate careers platform Bright Network. Chief executive Kirsten Barnes said AI is changing recruitment by helping identify authentic, high-potential candidates from a large pool in a fair, transparent, data-driven way. She emphasized the goal is to support, not replace, human judgment so teams can focus on potential instead of getting buried in volume.
Mishcon de Reya employs more than 1,400 people, including over 650 lawyers, with offices in London, Cambridge and Oxford, plus Singapore, Hong Kong and Riyadh.
What HR and Legal should scrutinize
- Lawful basis and DPIA: Confirm your lawful basis under UK GDPR and complete a Data Protection Impact Assessment before rollout. See the ICO's guidance on AI and data protection for benchmarks and controls. ICO guidance on AI and data protection
- Bias testing and monitoring: Run pre-deployment and ongoing adverse impact tests across protected characteristics. Document thresholds, fixes, and retests. Keep a human decision-maker accountable at each gate.
- Transparency and candidate rights: Provide a clear notice explaining what the chatbot does, data sources, retention, and how to request human review. Offer a non-AI route and reasonable adjustments for candidates who need them.
- Vendor governance: Lock down a data processing agreement, security posture (encryption, access controls, logging), and data residency terms. Check model versioning and update cadence.
- Interview quality: Maintain a governed question bank, scoring rubric, and escalation rules. Calibrate with human reviewers and run inter-rater reliability checks.
- Accuracy and integrity: Track false positives/negatives, drop-off rates, and candidate satisfaction. Add friction where needed (e.g., identity checks) without creating barriers that bias outcomes.
- Record-keeping: Define retention periods for transcripts and model outputs. Ensure audit trails are complete and discoverable.
Implementation checklist
- Complete DPIA and records of processing; brief your DPO and GC.
- Run a pilot with a small cohort; compare outcomes to human-led screening.
- Stand up a fairness dashboard with ongoing adverse impact metrics.
- Publish an AI-specific candidate privacy and transparency notice.
- Train recruiters on the scoring rubric, escalation, and when to override the AI.
- Offer an accommodations pathway and a human alternative at the same stage.
- Set data retention, redaction, and access policies for transcripts and logs.
- Review vendor SLAs, security attestations, and incident response plans.
Why this move matters
For firms processing thousands of applications, AI can free recruiters to spend more time on the candidates who stand out-while giving every applicant a chance to be heard. The win only holds if governance is tight: fair treatment, clear notices, and a human in the loop.
Practical next steps
- Map your current first-round process end-to-end; mark where AI helps and where humans must decide.
- Set success metrics before launch (time-to-screen, quality-of-hire proxies, adverse impact deltas).
- Pilot, measure, adjust. Keep a clear rollback plan if KPIs slip or risk rises.
If you're building a similar workflow, you may find these useful: AI Learning Path for Recruitment Coordinators and AI for Legal.
Your membership also unlocks:
