Co-creation in the age of Artificial Intelligence

As the world of AI continues to evolve, academics are facing a challenge in creating assignments that are useful for students in terms of both their academic study and developing skills, whilst minimising possible over-reliance on AI models to complete assessed work. 

Context 

Here at Manchester Metropolitan University, we teach very large cohorts on our Accounting and Finance degree. In our final year module on Corporate Reporting and Governance, we have over 300 students and previously have used case studies based on large multi-national listed entities as the basis for our assignment. The assignment itself is a business report, covering financial reporting and audit issues, looking at sustainability reporting, technology and business analysis. This approach had worked well for us for a number of years; however, the advent of AI has meant that we needed to rethink this. 

AI and assessment 

When basing an assignment on a large, listed company, students can use AI models to write much of their work. AI models contain information on listed companies and can quickly and easily construct an assignment which contains all the necessary points to achieve a very good grade. We put our previous year’s assignment into an AI model and with only around five minutes of prompting created an answer we believed would score more than 60% - which is a scary prospect! Therefore, we needed to look at how we could cover the same topics as in previous assignments, thereby meeting the desired learning outcomes, but ensure that AI is used responsibly and students still need to complete independent research. 

Assessment approach 

Private businesses are much less likely to be included in detail in AI models, and therefore we decided to contact a Chief Operating Officer (COO) of an organisation who was in our network, to ask if we could co-create an assignment brief. This was a preferred approach as due to our cohort size bespoke employer projects are a huge challenge, and we needed something that was scalable and that could be adopted quickly. 

Luckily for us the COO contacted was receptive. As joint module leaders, we wrote our desired assignment, focussing on this private company, and then asked the COO to review this to see if our focus areas aligned with his business. Once agreed, he sent us some information on his business that we could share with students, this did not contain any sensitive information but was something that couldn’t easily be found independently. We also referenced the company website, alongside databases our students have full access to, such as Statista and Mintel. We also asked the COO if he would be prepared to review the five best assignments received, and provide feedback for the students on this, which he agreed to. We then designed our assignment support tutorials and podcasts as normal and were ready to go! 

Benefits of the approach 

The first benefit we found was that students had to reduce reliance on AI models, as the level of detail required for the assignment was not contained within them. In tutorials and podcasts, we showed students outputs from AI, and explained that these were too vague, and did not contain the required detail specific to the organisation being used for the assignment. We do allow some AI use but ensuring that a student had to look at our suggested databases and the company information that was signposted meant that the level of independent research increased. 

Secondly, this assignment offered work integrated learning (WIL) to our students. They were working on an employer approved brief, with information provided by the employer and the possibility of employer feedback. This made the assignment further aligned with employability goals, by giving the students the chance to complete an assignment that was not just formulated by academics but had a ‘real-world’ employer focus. 

Results of the assignment 

We were concerned that the impact on our assignment marks may be very significant, and we did see a fall in the average mark from 66% in the 24/25 academic year to 61% in 25/26. We also saw a reduction in the number of firsts from 37% to 16% over the same period. However, this realignment was positive, as we could see that the highest scoring assignments were strong and showed excellent independent research, rather than rewarding those who had excellent AI skills. We also found that students engaged more clearly with the signposted online databases than in the prior year, helping them develop commercial skills. 

The five scripts that achieved the highest marks were sent to the employer for comments, and these comments were shared with the students involved. They were all really grateful to receive this feedback, which gave them confidence and something that they could reflect on and integrate into their own CVs etc when applying for graduate roles. This also ‘closed the loop’ on feedback. Although only a small number of students received personalised feedback from the employer, we still felt that this was beneficial and allowed students to see that their work could be reviewed in a commercial context, alongside an academic one. 

Learnings to take forward 

We would like to continue this model of assessment in our module, with the one major drawback being we need to look for a different company to base this on each year. We are looking for local small and medium sized entities (SMEs) that we could use for this, with the idea that we could have a ‘pool’ of employer contacts. We are also open to discussing with employers in more detail what areas they would like students to focus on, so we could end up with assignments that employers can draw on for their own business needs.  

We feel that this approach allowed us to mitigate some of the risk that AI poses, allowed students to develop commercial awareness and research skills, and gave them the opportunity to engage with a local company and potential employer. From the academic side, this was a fairly simple change to implement and importantly was scalable for large numbers. Co-creation really can be beneficial for the academic community, and we would urge others to consider this when designing assessments.