Jisc ran a year-long national pilot of AI marking tools (Graide, Keath, TeacherMatic) across 38 UK colleges and universities, finding real workload relief in formative assessment but persistent concerns over human oversight and student trust.
Jisc – National Centre for AI in Tertiary Education
Period
September 2025 – August 2026
Keywords
higher education, further education, assessment technology, edtech
Context
Jisc's National Centre for AI in Tertiary Education ran a year-long pilot (September 2025-August 2026) testing whether AI-powered marking and feedback tools could reduce assessment workload without compromising academic integrity, across 38 UK colleges and universities.
Activities
The pilot had two strands: three purpose-built education tools (Graide, Keath, TeacherMatic) and general-purpose assistants (ChatGPT, Gemini, Copilot). Institutions co-developed a shared student-communication pack and exchanged ethics-approval templates.
Results
Formative, low-stakes assessment emerged as the lowest-risk and most effective use case; maintaining genuine human oversight proved harder than expected, with parallel marking (AI and teacher marking side by side) identified as the most workable mitigation. Student reactions were mixed, with real concerns about trust and ethics.
Conclusions
Jisc's own framing is cautionary — 'AI should not be used as a catch-all' — and no quantified learning-outcome or workload-hours data has been published; the sector-wide evaluation stops short of claiming proven impact.
Implementation
Indicative cost
Medium (€50k–€500k)
Time to results
Short (< 1 year)
Staffing & skills
Jisc National Centre for AI in Tertiary Education (coordinator), 38 participating college/university staff
Conditions for success
Parallel marking (AI and teacher marking side by side) as the main human-oversight mitigation
Shared student-communication pack and exchanged ethics-approval templates across institutions
Common failure modes
No quantified learning-outcome or workload-hours data published
Mixed student trust in AI-marked assessment
Do you run this practice?
Claim it —
verified implementers get a public contact pathway and can propose corrections.
Data sources
Where this practice's information was retrieved from, and when.
A 12-week RCT with 190 Chinese university EFL students found Grammarly-based automated writing evaluation significantly improved task achievement, coherence, lexicon …