Overview

Cambridge Assessment English's Linguaskill is a certified English-language qualification used by universities, employers, and institutions worldwide. As international demand grew, Cambridge needed to move its delivery model: from a dependence on physical proctors at test centres to a technology-enabled approach that could support both in-centre and fully remote exam delivery across diverse geographies, without compromising the integrity standards a certified qualification demands.

The programme required direct integration with Cambridge's Metrica test delivery platform, a distributed network of 200+ exam centres spanning 22+ countries, and an operating model where every integrity decision remained in the hands of a trained human reviewer.

Cambridge selected Talview as its delivery partner.

Challenge

Cambridge faced a set of interconnected problems its existing model could not resolve at the scale required.

Physical proctors limited geographic reach. Expanding certified testing into new markets meant finding, training, and scheduling qualified invigilators in every location, a constraint that grew with every country added to the network.

Remote delivery introduced new integrity risks. Candidates sitting exams outside a controlled test centre bring variable devices, connectivity conditions, and physical environments. The exam integrity that a proctor in the room could observe now had to be documented by technology: not just flagged, but recorded in a form that holds up when a result is challenged.

Review demand outgrew review capacity. As remote and hybrid volumes scaled, the number of flagged sessions grew far faster than the pool of qualified human reviewers available to assess them. Without a way to close that gap, Cambridge faced a practical trade-off: let review turnaround slip, or hand integrity decisions to automation. Neither was acceptable for a certified qualification.

Integration had to be seamless. Any solution had to fit within Cambridge's existing Metrica candidate journey without introducing workflow fragmentation for administration teams operating across 22+ countries.

Solution

Talview designed and delivered a phased implementation built around three components: record-and-review proctoring, secure browser technology, and Alvy, Talview's AI proctoring agent, integrated directly into Cambridge's Metrica platform.

A two-phase rollout managed risk at each stage:

• Phase 1: Modernised in-centre delivery. Talview's proctoring and secure browser replaced the dependency on physical invigilators for session monitoring, while maintaining exam centre stability.
  
  
• Phase 2: Extended delivery to fully remote candidates, opening certified exam access beyond physical centres and supporting Cambridge's international expansion.
  

Alvy closed the review capacity gap.

Within the record-and-review workflow, Alvy sits between detection and the human reviewer. When a session is flagged, Alvy interprets the alert in context, examining identity verification status, the application in focus, any secondary device activity, and the timestamped screen and webcam evidence. It ranks flagged sessions by urgency and produces a reviewer-ready summary for each one.

The human-decision boundary is structural: Alvy triages and summarises, but every final integrity decision, and the reasoning behind it, is made and recorded by a trained human reviewer. Alvy never makes a disqualification decision.

Operational enablement underpinned the technology. Standardised training and playbooks for administrators, agents, and reviewers. Defined SLAs for review turnaround and results release. 24/7 candidate support and test-taking availability across all deployed regions and time zones.

Results 

Every figure below is from this deployment, under these conditions. None is generalised.

• 100,000+ candidates have completed Linguaskill exams through the Talview-powered delivery model.
  
  
• 22+ countries live, including India, Japan, Brazil, Germany, France, the USA, the UAE, and Peru.
  
  
• 200+ exam centres onboarded, with both in-centre and fully remote delivery modes operational.
  
  
• 4,000+ candidates in a single 12-hour window during university-led campaigns in Chennai, India and Peru, demonstrating sustained performance under peak load.
  
  
• Operational SLAs sustained at scale: review completion within 48 hours and results release within 5 days.
  
  
• Review completion within 48 hours and results release within 5 days, maintained at scale throughout the programme.
  
  
• 24/7 candidate support and test-taking availability across all geographies.
  

  
  Every integrity decision throughout has been made by a human reviewer.

WHAT THE RECORD LOOKS LIKE

Each reviewed session carries a timestamped evidence pack: the identity verification record, screen and webcam capture, a log of every flag with its context, and the reviewer's decision with their reasoning attached. That record is what Cambridge can put in front of a candidate, an institution, or an auditor if a result is ever challenged.

Conclusion

Cambridge's programme with Talview shows what it takes to move a certified qualification from physical proctors to remote delivery at international scale: technology that documents what is happening in the session, an AI triage layer that keeps reviewers focused on the sessions that need them, and a human-decision boundary that never moves.