AI in Assessments – What’s in Store for 2025

Dr. Rajeshwari Iyer examines how artificial intelligence is poised to reshape apprenticeship delivery and assessment as National Apprenticeship Week 2025 draws to a close. The transformation promises more efficient, personalised learning journeys for tomorrow’s workforce through AI-enabled feedback, analysis and skills verification.

As an AI EdTech developer, working with awarding bodies and end-point assessment organisations across the United Kingdom, I’ve been keenly observing their perspectives of using artificial intelligence in their assessment practices. As we begin 2025, here’s how I believe AI will reshape professional certifications and apprenticeship assessments, considering the implications for all stakeholders involved.

Current State and Imminent Changes

The overall assessment landscape is definitely experiencing significant transformation. Research by McKinsey (2023) indicated that 75% of education institutions planned to increase their AI investment in assessment technologies by 2025. Some current trends include Learning Management Systems and authoring tools integrating generative AI-based item generation & banking, data-driven predictive analytics tools for better performance insights and, more recently, marking & feedback of assessments.

While there is huge expectation and optimism about how AI will indeed contribute to better learner engagement, early adopters are taking a measured approach, as Deborah Hogget, Head of Product & Assessment at a leading STEM EPAO, SIAS, explains:  

“At SIAS, we recognise the transformative potential of AI in qualifications and End-Point Assessments. AI has huge potential, and by enhancing formative learning and personalising learner experiences, AI can enable innovative, learner-focused approaches to support flexible delivery and assessment models. However, to realise this potential, it is essential to balance this innovation with trust, rigour, and compliance, ensuring AI effectively supports and empowers learners and employers whilst benefiting the broader education sector.”

The ground truth remains that the TVET sector operates differently – while knowledge assessment forms a key part of teaching & learning, there is more emphasis on skills & behaviours – making assessors in the vocational sector still unsure of the real benefits of integrating AI within their existing workflows. Therefore, a good place to start is perhaps augmenting traditional knowledge assessments with AI-powered solutions for marking & feedback so that tutors can spend more time focusing on skills and behaviours.

As Kayle Brightwell, Director of Education, Safety Training Awards notes:

“We are excited about the potential AI has within assessment, not only making tutor time more efficient and reducing marking, but giving learners consistent and non-bias feedback / marks. AI doesn’t have personal or prior experience with a learners so marks their work according to the set mark scheme and is able to provide detailed specific feedback, helping tutors provide more personal learning to their learners. It is learning and developing all the time, which will create more opportunities for tutors to focus on and spend time with their learners.”

A key take-away from these conversations is that AI is here to stay with the promise of significant improvement in formative assessments with more timely feedback and personalised learning. The obvious place to start, which is already happening, is adopting AI to automate some of the marking, feedback & analytics that tutors need to do. This not only saves time for tutors but also provides detailed, high quality feedback on-demand to the learners, wherever they are, whenever they need it, as opposed to tutor time that happens once every 6-8 weeks. AI-assisted marking works well for text, math and audio based submissions and assessments, with further potential that can be unlocked using generative-AI.

So how about extending this to something more ambitious and aligned to the way this particular sector of education works?

Projected Scenarios for 2025

We all appreciate that apprenticeships and vocational qualifications exist because they allow for a more skills-based learning & assessment, rather than purely academic. This means trainees and apprentices can submit evidences of learning in any form of media – text, audio based reflections, images and even videos. Based on current technological trajectories and industry conversations, here are two illustrative scenarios of how AI might transform TVET assessment in 2025:

Scenario 1: Digital Marketing Apprenticeship

Drawing from emerging trends in marketing analytics, we might see AI tools that can analyse apprentices’ real-world application of skills by scanning social media campaigns, blog articles, and marketing analytics from ad accounts to evaluate key performance indicators, check for adherence to brand guidelines, and assess the application of marketing principles. Giving apprentices regular feedback on their actual work will keep them more engaged with their programme, uplift their confidence and ensure that e-portfolio submissions going to gateway are of highest possible quality.

Scenario 2: Installation and Maintenance Electrician (Level 3)

Building on existing video assessment capabilities and computer vision technology, we might see AI systems that can analyse video evidence of practical work submitted as part of e-portfolio. In this scenario, when apprentices submit recordings of tasks such as circuit installation or fault diagnosis, AI algorithms could:

• Track hand movements to ensure proper tool handling and technique
• Verify the correct sequence of safety checks
• Monitor the time taken for each step of the procedure
• Flag potential safety violations for assessor review

We can go a step further to instead of simply analysing content presented in the video, AI will extract competences evidenced in these multi-media submissions that can subsequently be matched to the knowledge, skills and behaviours listed in the standards for the qualification. While the AI wouldn’t make final assessment decisions, it could provide objective data to help human assessors give regular feedback and maintain consistency across multiple candidates. Moreover, knowing exactly what KSBs have been met and where gaps in learning exist will allow for more meaningful tutor time with targeted learning journeys.

These scenarios represent potential applications of AI in apprenticeship assessment, but can be extended to any vocational qualification.

We Need to Act Now

The TVET sector is growing and is ripe for innovation – it involves learning highly specific skills, working in high-stakes roles such as Engineering, healthcare, energy and business administration, in collaboration with some of the largest companies in the world – organisations that are developing and using AI for massive efficiency gains, for better productivity, for employee professional development, so why should the sector be left behind? Senior leaders in the sector need to begin with AI at the core of any strategy – allocate time and resources to upskill themselves and their staff, develop policies for safe use of AI, and ensure smooth digital transformation before AI transformation. AI is not just large language models or chatGPT; there are different kinds of AI built for specific use cases, so it’s important to take a stepped approach to identify your organisation’s needs and challenges first before adopting the right tool.

By Dr. Rajeshwari Iyer, Co-Founder & CEO at sAInaptic