Embracing AI effectively for assessment and to develop employability skills

For the last three years at University of Sussex Business School, we have taken the view that generative AI is here to stay and have been evolving our approach to attempt to use it productively, rather than treating it only as a threat to academic integrity.

What we’ve arrived at this academic year is by no means final. It will continue to be shaped by research, technology and employer needs, but we feel it is an assessment approach that is both realistically “AI-saturated” and strongly aligned with the analytical and critical skills employers seek in graduates.

Our community of accounting educators gets stronger the more we work collaboratively. What follows is an outline of our approach on two modules. We welcome comments and discussion and of course hearing about other approaches.

The assessments

AI-supported analysis and evaluation of accounting concepts in practice

Alongside an unseen exam needed to meet accreditation we have a 25% weighting for Advanced Financial Accounting coursework at level 6 (final year of an undergraduate bachelor’s degree). The assignment required students to consider three sets of published financial statements, the IFRS conceptual framework, and specific accounting standards, all in order to critically evaluate the extent to which the stated aims and objectives of those standards appeared to have been met in practice.

In the olden days (2023…), students tackled similar analytical work manually; in 2025/26 the use of AI tools was mandatory. Without these tools the sheer volume of data, disclosure detail, and cross-referencing between the standards and the company reports, would have been unmanageable and unfair for just 25% of the overall grade. By mandating the use of AI tools, students worked in a way that much more closely resembles real professional practice. The assignment was not to get students to simply have AI do the work, students needed to develop and demonstrate an ability to use AI tools to process large amounts of source data in a meaningful way.

AI-assisted balanced scorecard analysis of companies

A similar component in Advanced Management Accounting (40% essay) required students to evaluate the use of balanced scorecard concepts within three companies and critically consider how they fit with their adopted strategies and accountability practices. Students were expected to use a generative AI tool as an assistant to summarise the companies’ balanced scorecard concepts present in annual reports. Prior to 2024, students would have been required to evaluate the balanced scorecard of just one company, given that they will need to read through a vast number of pages of their annual report to identify its elements: objectives, planned measures, results, targets, etc. and then evaluate them. The use of AI for document summarisation helped to streamline this process, allowing students to analyse a larger number of companies and enabling us to place greater emphasis on higher-level critical analysis rather than on simply identifying information already contained in annual reports. While AI provided access to greater volumes of information to support decision-making, its effective use and the critical evaluation of its outputs, alongside the other requirements of the essay, were intended to help develop students’ AI literacy skills in this area.

How students used AI, instructions given and where the real value is obtained

In seminars designed to develop skills needed for the assessment, students were instructed to use AI tools to consider how the implementation of IFRS 18 Presentation and Disclosure in Financial StatementsInternational Accounting Standards Board (2024). IFRS 18 Presentation and Disclosure in Financial Statements. London: IFRS Foundation. (Accessed: 24 June 2026). in future years would change the presentation of a current set of financial statements. Through this exercise of groups using the same sources, we were able to consider and explore how the outputs of the AI tools differ, and how best to interrogate the outputs to arrive at conclusions students were confident of.

The exercise was also a powerful lesson in how quickly we can all navigate lengthy annual reports, extract specific disclosures, and map the same back to particular IFRS requirements. The real value of this comes from guiding students away from accepting AI outputs at face value, and to apply critical thinking and broader learning to shape and interrogate those outputs, the same skills being assessed in the coursework.

As a result, strong students were able to develop their ability to craft precise, iterative prompts rather than broad “do my assignment” requests and verify AI-generated interpretations against the actual wording of IFRS and the annual reports. It was quite easy to differentiate between students, based on their ability to make deliberate decisions about which AI-generated ideas merited inclusion, adaptation or rejection in their final work. Of course, some students will have used AI to help with this also, but that is arguably another strength of using the tools effectively.

In preparation for the second assignment, we dedicated a seminar to improving students’ use of a generative AI tool. Students were provided with a simple prompt used to generate a summary of the balanced scorecard of three companies, along with the output. We asked students to:

• evaluate the output produced by the prompt, specifically assessing its usefulness for meeting the coursework requirements;
• revise the prompt using the same annual reports in the recommended generative AI tool until they were satisfied with the output (covering both financial and non-financial performance measures);
• evaluate aspects of the revised balanced scorecard summary of each company that align with its strategy, and use selected elements to assess how accountability is demonstrated in those companies.

Following the seminar, we offered students individual informal feedback on their prompts and outputs for their chosen companies intended to be used for the actual essay.

Students were required to include, as an appendix, the conversations they had with AI tools throughout the assessments. This served several purposes:

1. It made AI use transparent, rather than something students felt they had to hide.
2. It allowed markers to see how students iterated their prompts and responded to outputs.
3. It offered rich evidence for assessing critical engagement with AI, not just the polished end product.

Both assessments received grades based on five standard perspectives for our level 6 students (knowledge and understanding, application, critical thinking, reading and research, and presentation and style). However, this was informed not just from the final essay, but the appendix that showed the interaction and use of AI. Students who challenged, refined and tested AI outputs, demonstrating deeper understanding of both the financial reports and the relevant standards and theories, as a result scored much higher than those taking AI outputs at face value.

For students and markers this process enabled a much more explicit demonstration of critical thinking and research compared to a traditional essay. The essay itself was still the main driver of the final grade, but the cognitive process behind the work was displayed.

Outcomes compared with pre-AI assessments

The overall distribution of marks looks remarkably similar to essay-style assessments from pre-AI cohorts for both modules, even though the quality and depth of even mediocre submissions were far greater in 2026 than an excellent submission in 2023. This is the power of embracing AI in education and the workplace – we are able to achieve so much more in the time available, and the marking guidance took this into account, raising the expectations of acceptable, good, and excellent work.

Students who invested minimal effort still tended to submit work that meets some of the basic requirements but remained descriptive and superficial (similar to the pre-AI period, these students were rewarded with lower marks). Those who engaged actively with both the AI and the underlying technical material, and who applied concepts learned throughout the module, produced much richer, more analytical work, and were rewarded accordingly. Most fell somewhere in between. The mandated use of AI ensured fairness across the cohort; we did not see students who submit ‘authentic’ human work score lower than students submitting AI supported work.

Using AI explicitly in the assessment design and teaching delivery, at least in this context, does not inflate grades or remove discrimination between levels of performance. Instead, it shifts the focus of what was being discriminated: away from the ability to manually trawl through documents, and towards the ability to direct tools and exercise critical thinking and judgement over their outputs.

Signals from employers and former students

One particularly striking piece of feedback came indirectly, via a student conversation with their new employer. After the student described the assessment – especially the requirement to work with AI tools to interrogate financial statements and standards – the employer’s reaction was incredibly enthusiastic. Their response, paraphrased, was that this was “exactly the kind of capability they were looking for in graduate hires”.

A similar view was expressed during a guest lecture from a former student, now a Senior Finance Business Partner, discussing with students how they use AI in day-to-day management accounting tasks, particularly as a ‘thought partner’. This real-world usage underlined the importance of embracing AI tools in work, study and assessment, rather than prohibiting its use.

Although anecdotal, these reactions resonate with wider conversations about employability in accounting: employers need graduates who can combine technical knowledge with tool-use, scepticism and communication, rather than relying solely on manual processing.

Reflections and next steps

For us, these assessments have reinforced the idea that generative AI can be built into accounting education in ways that strengthen, rather than undermine, core learning outcomes. By requiring students to use AI openly, and by assessing how they guide and evaluate its outputs, we can:

• Preserve academic standards and discrimination between levels of performance.
• Develop students’ ability to work with, not against, the technologies they will encounter in practice.
• Surface and evidence employability skills that might otherwise remain implicit.

There is still much to explore: equity of access to tools, the scalability of marking AI appendices in large cohorts, and the best ways to integrate explicit conversations about ethics and professional responsibility. But this experience suggests that embracing AI in assessment design, rather than ignoring it, can bring us closer to the realities of contemporary accounting work.

If any accounting educators are interested in discussing the approach above, or indeed sharing their own approaches please contact us.

Dr Oluwaseun Osituyo

Dr Oluwaseun (Seun) Osituyo is an Associate Professor of Accounting and the Faculty Student Experience Lead at University of Sussex Business School. She is also the Director of MSc Accounting and Finance course at Sussex and leads undergraduate and postgraduate accounting modules. Seun is a Senior Fellow of the Higher Education Academy.

Her teaching and research are driven by a commitment to fostering inclusion, innovation, and active engagement. She is also a member of the Institute of Chartered Accountants of Nigeria (ICAN). Before joining academia in 2017, she had practical experience in various accounting roles in Nigeria.

Contact: O.Osituyo@sussex.ac.uk

Matthew Walsh

Matthew Walsh is a Chartered Management Accountant and Associate Professor in Accountancy at the University of Sussex. As well as convening accounting modules at postgraduate and undergraduate level, he is also the Deputy Head of Department of Accounting and Finance.

Previous roles include Education and Student lead and course director. Alongside academia, he remains a practicing management and tax accountant with 30 years’ experience that spans a wide range of industries, for and not-for-profit, large and small. This ongoing work outside of teaching heavily informs current teaching on management accounting, taxation, and accounting in practice.

Contact: M.Walsh@sussex.ac.uk

Part of the Pedagogy series
Join the Accounting Cafe community
© Accounting Cafe

How to cite this article: Osituyo, O. & Walsh, M. (2026) ‘Embracing AI effectively to assess final-year accounting students and develop skills essential for employment’, Accounting Cafe, 22 June. Available at: https://accountingcafe.org/2026/06/22/ai-in-accounting-assessment/ (Accessed: [insert date])