Five years of the Assessment Innovation Fund: AI, Inclusion and Trust Part 2

To continue the camping theme from part one in this series, even if you think you have all the equipment and have predicted all the scenarios, a storm can come along and completely change the picture. That storm was generative AI. Suddenly, the questions being asked about assessment became much bigger.

AI was present in the early windows, but it had not yet become the force reshaping assessment conversations. In the Assessment Innovation Fund (AIF) register, around 19% of application entries across funding windows one and two referenced AI, automation, machine learning, or adaptive approaches in some form. Across windows five to seven, that rose to around 44%.

What’s most interesting though is that applicants were rarely focussed only on replacing human judgement. Instead, they were exploring where AI could add value like providing quicker feedback, reducing administrative burden, identifying learner needs earlier, supporting inclusion, personalising learning, and helping tutors focus their time where it matters most. The Open University’s AIF-funded research shows why this matters. Its project looked at generative AI and assessment robustness across 944 scripts, 59 questions, 17 assessment types and 43 markers. AI-generated answers achieved passing marks on 58 of the 59 questions tested. Just as importantly, training markers to identify AI-generated work improved detection, but also increased false positives, meaning genuine learner work was more likely to be wrongly suspected of being AI-generated.

That’s not just a technical problem. It’s a fairness problem.

Some of the so-called “hallmarks” of AI, such as generic language, overly formal writing or shallowness can also appear in the work of learners who need more support, or learners writing in an additional language.

So, the lesson here is not “panic and buy a detector”. It’s that AI should push us to ask better questions about assessment design. Are we assessing the right things? Is our marking guidance clear enough? Are we supporting learners and tutors to use AI ethically and confidently?

It turns out the answer to AI in assessment might be less “catch them out” and more “design it better”.

Who gets to take part?

The most recent applications suggest another shift too. Assessment innovation is becoming much more about confidence, inclusion and progression.

Inclusion was present in earlier windows, but it becomes far more visible later. Around 9% of entries across windows one and two referenced inclusion or accessibility-related themes, compared with around 23% across windows five to seven. Later applications include more direct references to ESOL, SEND, neurodiversity, dyslexia, dyscalculia, accessibility, anxiety and underserved learners.

It suggests the sector is no longer thinking about assessment innovation only as a productivity or technology issue. Increasingly, it is being seen as a participation issue – who can access assessment, who feels confident demonstrating their skills, and who gets the opportunity to succeed. Bodyswaps’s in-flight project is a good example of this newer phase of thinking. The project is exploring how immersive technology and AI can support vocational learners preparing for end-point assessment, with completion due in July 2027.

Its first phase feasibility report has already provided useful early learning. Learners who completed the pathway reported improvements in confidence and communication skills, and tutors saw particular value in the professional discussion simulation. But the first phase also showed why real-world testing matters. The pathway needed refinement; it was too long,

completion rates were lower than expected, and engagement was strongest when activities felt directly relevant to learners’ occupational goals.

Innovation is not always about discovering a completely new route. Sometimes it’s about improving the path that already exists. Sometimes, it means finding out that the thing you built is too long, needs more vocational relevance, and works best when tutors are actively involved.

Less glamorous, maybe. More useful, definitely.

Make the learning travel further

Looking back across the fund, there is a clear journey.

At first, many ideas were about making assessment digital. Then the focus moved towards making assessment more authentic. More recently, applications have increasingly focussed on AI, inclusion, confidence, feedback, and trust.

That evolution tells us something important about the sector.

People are not innovating for the sake of it. They’re responding to real pressures including learner confidence, tutor workload, employer relevance, assessment integrity, AI disruption, inclusion, and the need to recognise a broader range of skills.

The challenge now is that too much of this learning remains hidden in individual projects, research reports, and small communities of practice, which feels like a missed opportunity.

Nobody comes back from a camping trip and says, “I found the best picnic spot with a riverside walk and a café with great coffee, but I’ve decided to keep it all to myself.” At least, they shouldn’t.

If we want assessment to evolve, we need to make the learning from that work easier to find, understand, and apply. Not just the polished final version, but the practical learning too. What motivated people to apply? What problems were they trying to solve? What worked? What didn’t? What needs more evidence? The Assessment Innovation Fund has shown that the sector does not lack ideas. What we need now is better visibility of those ideas and a stronger shared understanding of what they are teaching us.

Because the future of assessment will not be shaped by one platform, one funding round, or one policy announcement. It will be shaped by the accumulation of learning from real projects, tested with real learners and tutors, in real settings.

That is why innovation in this space matters. Not because every idea will succeed, but because every tested idea can help the sector move forward.

By Dean Blewitt, Senior Innovation and Investments Manager at NCFE