Strategy & Leadership
Building the Business Case for AI
Assessment at Your Institution
Marking loads are unsustainable, NSS scores stall, and student
expectations are rising. Here is how to frame AI assessment as a
quantifiable case for a senior team or finance committee.
Eduface
·
10 min read
·
Written for DVCs & finance leads
You can see the case for AI assessment clearly enough from where you sit. The marking
burden is unsustainable, NSS scores in Assessment and Feedback remain stubbornly low,
and students are waiting weeks for feedback that arrives after the learning moment has
passed. The question is not whether to act. The question is how to build a case that
moves a senior team or finance committee from "interesting idea" to approved budget
line.
How do you build a business case for AI assessment in higher education?
The business case for AI assessment rests on three measurable pillars: reducing the
cost and volume of marking time per assignment, improving NSS scores in
Assessment and Feedback (the sector's most consistently low-scoring category), and
managing compliance risk under the EU AI Act. Each pillar has quantifiable inputs that
finance teams recognise, and all three connect to strategic priorities already on senior
leadership agendas.
What is the actual cost problem that AI assessment solves?
The most direct line from AI assessment to a finance committee is time. Marking is the
single most labour-intensive academic activity outside contact hours, and it is largely
invisible in institutional cost modelling because it is absorbed into salaried lecturer time
rather than tracked as a discrete cost.
Research from the University and College Union (2016) found that 26% of higher
education staff in the UK work more than 50 hours per week, that 75% describe their job
as stressful, and that 46% report unrealistic time pressures as a chronic feature of their
role.
1
Marking and feedback account for a significant share of that overload, particularly in
departments with high-volume written assessment.
26%
of UK HE staff work
more than 50 hours
per week
75%
describe their job as
stressful
95%
Eduface AI
alignment with
lecturer marks in UK
pilots
3 wks
average NSS-cited
wait for meaningful
feedback
A practical starting point for any business case is the marking audit: how many written
submissions does your institution process per academic year? What is the average time
per submission, including first marking, moderation, and feedback writing? At a modest
estimate of 45 minutes per submission across 30,000 annual submissions, that is 22,500
hours of academic staff time. Reducing that figure by even 30% through AI-assisted
marking and automated feedback generation represents a material staff cost recovery.
How does AI assessment connect to NSS performance?
The National Student Survey has five questions dedicated to Assessment and Feedback,
covering criteria clarity, feedback usefulness, timeliness, help in understanding
performance, and fairness. Assessment and Feedback is the lowest-scoring category in
the NSS year after year, and it is the category that has the clearest connection to both
student retention and the Teaching Excellence Framework.
The mechanism between AI assessment and NSS scores is direct. Hattie and Timperley
(2007) established that the most effective feedback is specific to the task, delivered while
the student can still act on it, and tied to explicit learning goals.
2
AI assessment tools
deliver on all three: feedback is criterion-referenced, turnaround drops from weeks to
days, and every student receives the same structured, detailed response rather than a
comment that varies by how many papers remain in the stack.
Nicol and Macfarlane-Dick (2006) identified seven principles of good feedback practice,
including that feedback should help students self-assess and should be delivered in time
to influence further learning.
3
Institutions with consistently low NSS scores on feedback
timeliness should be able to demonstrate a direct line from AI assessment deployment to
improvement on those specific survey questions.
Three Pillars of the AI Assessment Business Case
⏱
Staff Time
Reduce marking hours
per submission
Input: submissions/yr
× avg. minutes/sub
× staff hourly rate
Quantifiable cost saving
📊
NSS Performance
Improve Assessment
& Feedback scores
Faster feedback
More consistent marking
Criterion-referenced detail
TEF and reputation impact
⚖
Compliance Risk
Manage EU AI Act
obligations proactively
Avoid enforcement risk
Student trust & transparency
Governance framework
Risk mitigation value
Figure 1: The three business case pillars for AI assessment in higher education. Each can be quantified in terms
meaningful to a finance committee or senior leadership team.
How do you frame the return on investment?
The honest answer is that calculating a precise ROI for educational technology is
complex, and any business case that overpromises should be treated with scepticism.
What is practical is building three cost categories, each of which stands independently.
Direct cost recovery
AI assessment tools reduce the time lecturers spend on first-pass marking and feedback
writing. Institutions can calculate this by surveying average marking time per submission
type, then modelling the time saving at a realistic reduction figure (30 to 50% is a credible
range based on current pilot data). This time does not disappear from the institution: it is
recovered for student contact, research, or administration that currently gets
deprioritised.
Quality improvement value
NSS scores in Assessment and Feedback have a direct relationship with institutional
reputation, league table position, and student recruitment. A one-point improvement in
satisfaction scores in this category, sustained over two or three years, translates to
measurable recruitment and retention effects. This is harder to quantify precisely but is a
legitimate strategic benefit that resonates with DVCs and Pro Vice-Chancellors for
Education.
Compliance cost avoidance
The EU AI Act creates obligations for institutions deploying AI in assessment from August
2026. Institutions that deploy unreviewed or non-transparent AI systems face
enforcement risk. Procuring a compliant tool now avoids the cost of remediation later, and
positions the institution as a responsible adopter rather than a latecomer forced into
compliance.
What objections should the business case address?
Senior leadership teams and finance committees will raise predictable objections. A
strong business case addresses them directly.
Objection
Response
"AI can't match
human judgement."
In Eduface's UK pilots, AI marks aligned with lecturer marks 95% of
the time. The system flags divergence for closer human review. The
lecturer retains final authority over every grade.
"Students will be
unhappy with AI
marking."
Student dissatisfaction is more strongly linked to slow, generic
feedback than to who produced it. NSS data consistently shows
timeliness and specificity of feedback as the core pain points, both
of which AI assessment improves.
"It will take years to
implement."
Eduface integrates with Blackboard, Brightspace, Moodle, and
Canvas. A departmental pilot can be running within a semester with
no new infrastructure required. Institutions on the Jisc/CHEST
framework can procure without a full tender process.
"How do we know it's
legally compliant?"
Eduface is designed around the human-in-the-loop principle
required by Article 14 of the EU AI Act. Human sign-off is mandatory
before any grade is released. Processing runs on EU-based
infrastructure without third-party API dependency.
"What's the evidence
from comparable
institutions?"
Eduface has pilot partnerships at Bath Spa University in the UK and
De Haagse Hogeschool, Tilburg University, and Hogeschool
Rotterdam in the Netherlands, with verified accuracy and workload
data available on request.
How should a pilot be structured to build internal evidence?
The most effective approach to internal advocacy is a controlled departmental pilot that
generates your own data rather than relying solely on external evidence. A well-designed
pilot over one semester can answer the questions a senior team will ask before
committing to institution-wide deployment.
A useful pilot design would cover at least two modules with different assessment types,
measure lecturer time spent on marking before and after AI assistance, collect student
satisfaction data on feedback quality and timeliness at module level, and compare marks
on a sample of submissions to verify alignment. After one semester, you have real data
from your own institution, in your own disciplinary context, with your own student cohort.
Recommended Pilot Timeline (one semester)
Weeks 1–2
Setup: rubric
configuration
Weeks 3–4
First submissions
marked with AI
Mid-semester
Time tracking
review checkpoint
End of semester
Data review: time,
accuracy, satisfaction
Post-pilot
Report to senior
leadership team
Figure 2: A recommended one-semester pilot structure. Measuring time, accuracy, and student satisfaction at
module level produces institutional evidence for a senior leadership decision on wider deployment.
Procurement note: UK institutions on the Jisc/CHEST framework can procure
Eduface without running a full tender process. Irish institutions can procure via
HEAnet. Contact Eduface for a framework pricing overview and pilot terms.
Who needs to be in the room when the business case is made?
The conversation about AI assessment typically starts in one place and needs to land in
another. A Learning Technologist or Director of Learning and Teaching may be the
champion, but the decision to fund and deploy institution-wide requires sign-off from the
Deputy or Pro Vice-Chancellor for Education, often with Finance and Legal involved on
cost and compliance questions respectively.
Building the business case means preparing materials for each of those audiences. For
the PVC-Education: NSS impact, TEF alignment, and student outcome evidence. For
Finance: a time-and-cost model based on real submission volumes. For Legal or
Compliance: the EU AI Act compliance position and the documentation that comes with an
approved framework supplier. Bringing all three to the same conversation, with the same
pilot data as the shared evidential base, is where the decision moves from deferred to
approved.
Frequently asked questions
How long does it take to implement Eduface at an institutional level?
A departmental pilot can be active within a few weeks of procurement, given Eduface's
native integrations with Blackboard, Brightspace, Moodle, and Canvas. Institution-wide
deployment typically follows a phased approach over one to two semesters, beginning
with volunteer departments and expanding on the basis of pilot results. UK institutions on
the Jisc/CHEST framework can begin procurement immediately.
What data does a pilot produce that is useful for a senior leadership decision?
The most compelling pilot metrics are: lecturer marking time per submission before and
after AI assistance (measurable via simple time-logging); AI-to-lecturer mark alignment
rate for the pilot cohort; student satisfaction scores on feedback timeliness and specificity
collected at module level; and any qualitative feedback from lecturers on the workflow.
One semester of this data, from two or three modules, provides a credible evidence base
for a wider deployment decision.
Can AI assessment help with the consistency problem as well as the speed
problem?
Yes, and this is often the stronger argument with academic quality committees. Human
marking of the same rubric by different markers produces measurable variation: research
consistently shows that general impression marking in particular is unreliable. AI marking
applied to the same rubric is perfectly consistent across all submissions in a cohort. This
reduces the need for extensive moderation, which itself represents a significant time
saving on top of the direct marking reduction.
Does the business case apply to smaller institutions or only large universities?
The business case scales with submission volume but applies at any size. Smaller
institutions with fewer marking hours may weight the NSS and compliance arguments
more heavily than the direct cost recovery. For institutions with limited staff capacity, the
opportunity to redirect time from first-pass marking to student contact and support is often
the most compelling argument regardless of the absolute hour count.
Start with the evidence you can produce yourself
The strongest business case is one built on your institution's own data rather than
external studies alone. A one-semester pilot, designed to measure the right outcomes,
gives a senior leadership team the confidence to move from experiment to investment.
The cost of a pilot is marginal. The cost of another year of declining NSS scores and
unsustainable marking workloads is not.
Start with a pilot
Request a demo and ask about Eduface's pilot programme for
UK and Irish institutions. Eduface is approved on the Jisc/CHEST
and HEAnet frameworks, so procurement is straightforward.
Request a demo
References
University and College Union. (2016). Workload is an education issue: UCU workload survey report 2016.
UCU. [Key findings: 26% of HE staff work 50+ hours/week; 75% describe their job as stressful; 46%
report unrealistic time pressures.]
Hattie, J., & Timperley, H. (2007). The power of feedback. Review of Educational Research, 77(1), 81–112.
Nicol, D. J., & Macfarlane-Dick, D. (2006). Formative assessment and self-regulated learning: A model
and seven principles of good feedback practice. Studies in Higher Education, 31(2), 199–218.
Bloxham, S. (2009). Marking and moderation in the UK: false assumptions and wasted resources.
Assessment & Evaluation in Higher Education, 34(2), 209–220.
Black, P., & Wiliam, D. (1998). Assessment and classroom learning. Assessment in Education: Principles,
Policy & Practice, 5(1), 7–74.