Dr Salman Shahid is a Senior Lecturer (Teaching and Scholarship) in the Department of Chemical Engineering at the University of Manchester. He also serves as the Discipline Head of Education for Chemical Engineering. His work focuses on enhancing inclusive assessment, feedback, and student learning through the thoughtful integration of artificial intelligence both inside and outside the classroom. His current interests include AI-enabled feedback, assessment innovation, student partnership, and the development of evidence-informed approaches to Generative AI in higher education.

 

Artificial Intelligence (AI) is rapidly becoming part of everyday learning. Students are increasingly using tools such as Microsoft Copilot to explain concepts, generate ideas, revise for assessments, analyse information and receive feedback. At the University of Manchester, these conversations are particularly timely as the University continues to develop its institutional approach to AI and expands access to Microsoft 365 Copilot across the University community. The University has emphasised both the opportunities and the risks of AI while encouraging its responsible and effective use^[1-2].

Much of the discussion in higher education has understandably focused on AI and assessment. Questions of academic integrity, assessment design and appropriate use remain critically important. However, there is another question that deserves equal attention:

What does effective AI-supported learning actually look like for students?

Many students now have access to powerful AI tools. What they often lack is practical guidance on how to use those tools in ways that support learning rather than replace it. While policies help students understand what is permitted, they do not necessarily help them understand how AI can be used to develop critical thinking, judgement and independence.

To help address this challenge, I have developed an AI and Microsoft 365 Copilot Toolkit for University Students, which accompanies this article. The toolkit provides practical learning activities, Socratic prompting strategies, discipline-specific examples, verification checklists and guidance on avoiding over-reliance. Although initially developed within Chemical Engineering, it has been designed to be adaptable across disciplines and educational contexts.

 

Downloadable Copilot toolkit:

AI_Copilot_Toolkit_All_University_Students.docx

From using AI to learning with AI

A student struggling with a problem might ask Copilot:

“Solve this problem.”

The response may be impressive and even correct. However, the student may learn very little from the interaction. Alternatively, the student might ask:

“Don’t solve this problem. Ask me one question at a time to help me work it out myself. Challenge my assumptions and help me identify mistakes.”

The difference is subtle but important.

The first approach risks outsourcing thinking.

The second approach uses AI to scaffold thinking.

Students still need to analyse information, justify decisions, evaluate evidence and reflect on their understanding. AI becomes a learning partner rather than a substitute for learning.

This principle applies across disciplines. An engineering student can use Copilot to challenge a design decision. A humanities student can use it as a critical discussion partner. A business student can use it to interrogate assumptions behind a strategy. A healthcare student can use it to justify competing interpretations of a case.

The discipline changes. The underlying pedagogy does not.

A simple framework for AI-supported learning

The toolkit is built around a six-stage learning cycle:

THINK → ASK → CHALLENGE → APPLY → VERIFY → REFLECT

Students are encouraged to:

• Think before using AI and attempt the task themselves.
• Ask clear questions with appropriate context.
• Challenge AI outputs by questioning assumptions and evidence.
• Apply what they have learned independently.
• Verify information, calculations, sources and reasoning.
• Reflect on what they can now do without AI support.

This final stage is particularly important.

The aim of AI-supported learning should be greater independence, not greater dependence.

The toolkit provides practical examples of how students can use this framework in revision, problem solving, essay writing, project work, data analysis and research activities.

 

Copilot as a Socratic tutor

One of the most powerful applications of Copilot is also one of the simplest.

Students can explicitly ask it not to provide answers.

For example:

“Act as my Socratic tutor. Ask one question at a time. Do not give me the answer. Challenge my assumptions and help me identify my own mistakes.”

This transforms the interaction.

Instead of generating solutions, Copilot generates questions.

Instead of reducing cognitive effort, it redirects cognitive effort.

This approach can be used for revision, problem solving, oral examination preparation, case-based learning, laboratory reflection, research projects and dissertation supervision.

For educators, it also provides opportunities to design AI-supported learning activities that focus on questioning, critique and reflection rather than answer generation.

Supporting higher-order thinking

One concern frequently raised about generative AI is that it may encourage superficial learning.

However, this depends largely on how the technology is used.

The toolkit encourages students to use AI to support higher-order cognitive activities by asking questions such as:

• What assumptions am I making?
• What evidence would challenge this conclusion?
• What is the strongest argument against my interpretation?
• What alternative explanations should I consider?
• Which part of my reasoning is least secure?

These questions move students beyond information retrieval and towards analysis, evaluation, synthesis and reflection.

Used in this way, AI can support the very capabilities that higher education seeks to develop.

 

Tackling over-dependence

At the same time, AI literacy must include understanding when not to use AI.

Many students value AI tools but also express concerns about becoming over-reliant on them. Developing AI fluency therefore requires building habits that promote independent learning.

The toolkit encourages students to:

• Attempt tasks before using AI.
• Ask for hints rather than complete solutions.
• Use Socratic questioning approaches.
• Maintain regular AI-free study activities.
• Verify all important outputs.
• Reflect on what they have learned.
• Gradually reduce AI support as competence develops.

This progression can be viewed as an independence ladder:

Explain → Hint → Question → Challenge → Critique → Independent Performance

The objective is always to help students move towards independent judgement and problem solving.

From technology adoption to pedagogical design

The arrival of Copilot provides an opportunity to move beyond asking whether AI should be used in higher education.

Instead, we can ask:

What kinds of learning do we want AI to enable?

The technology itself is not the pedagogy.

A prompt is not automatically a learning activity. The educational value comes from what students are required to think about, apply, justify, evaluate and reflect upon.

For this reason, I see the accompanying toolkit not as a collection of “best prompts”, but as a collection of learning strategies supported by AI. It is intended as a practical resource that programme teams, lecturers and students can adapt to their own disciplinary contexts.

 

An invitation to experiment

The next step is not to create a perfect AI toolkit.

It is to use it.

Try one AI-supported activity in a seminar. Ask students to use Socratic mode during revision. Give them an AI-generated answer and ask them to critique it. Invite them to reflect on when AI helped their learning and when it hindered it.

The toolkit has been developed as a starting point for those conversations.

Ultimately, the goal is not to produce graduates who can simply ask AI for answers.

The goal is to develop graduates who can question those answers, evaluate evidence, challenge assumptions, make informed decisions and know when they need to think for themselves.

That is what meaningful AI fluency should look like.

The opportunity

The arrival of Copilot gives us an opportunity to do more than teach students how to use another digital tool.

We can teach them how to think with AI without letting AI think for them.

For Chemical Engineering, this means moving from:

“Can AI solve this engineering problem?”

to:

“Can the student use AI to become better at solving engineering problems?”

That is a much more interesting educational question.

And perhaps the ultimate measure of success is not how frequently a student uses Copilot, but whether, after using it, they are able to close the laptop and solve the next problem themselves.

For more detailed Discipline specific (Chemical Engineering) AI use in teaching and learning guideline please see the resource: Use of AI Guidelines for Students and Staff Chemical Engineering

 

Resources:

1. University AI Guidelines: https://www.staffnet.manchester.ac.uk/ai-hub/ai-guidelines-and-policies/ai-guidelines-overview/guidelines/
2. Copilot information: https://www.itservices.manchester.ac.uk/ourservices/microsoft365/copilot/
3. AI in Teaching and Learning blog: https://blogs.manchester.ac.uk/itl/category/ai/

 

 

Related