AI agents: Test your skills in real-life scenarios

With AI agents students to practise and apply the knowledge acquired during a course. Scenario or simulation agents can be utilised to simulate real-life situations encountered in working life. The agent is a tireless training partner, acting as a tutor: asking additional questions, challenging, commenting, and simulating authentic scenarios in which the student must use their learning. This practice closely resembles the real-life situations students will face after graduation. AI agents do not replace teachers but are an excellent addition to the learning toolkit. Learning begins with assimilation and deepens through application.

How can a scenario agent be used in a course?

The agent could be used in a course, for example, as follows. First, the student studies the course materials and assimilates the topics to be learned, participating in teaching, making use of materials, and completing assignments. Then, the student proceeds to converse with the Copilot agent, which is trained to act in a specific role according to the course topic, such as a customer, supervisor, patient, or project stakeholder. The agent, however, does not provide ready-made answers.

The agent has several advantages over traditional practice methods:

• The agent is always available: The agent does not tire, rush, or restrict practice times. The student can return to the conversation at any time according to their own schedule.
• A safe environment to make mistakes: Many students hesitate to ask “stupid questions” or practice situations that make them nervous. The agent creates an environment where a culture of trial and error is a natural part of learning.
• Simulation of real-life situations: The agent adopts the given role and, depending on the course objectives, can act as a challenging customer, patient expert, novice colleague, another student, or an empathetic patient.

The agent does not give ready-made answers. An important part of learning is testing one’s own understanding. The agent is specifically trained to act as a tutor. Therefore, the agent offers a limited and controlled way to utilise AI even in those courses and tasks where free use of AI is not desirable. Agent also introduces students to a useful way to use AI to support their learning. The agent asks follow-up questions, requests clarifications, challenges the student to justify their solutions, and simulates realistic, unclear, or contradictory situations that occur in working life. This strengthens the student’s ability to apply theory in practice.

Using the agent is resembles of the Flipped Classroom approach, as it emphasises that learning happens best when students actively apply the knowledge they have assimilated. The AI agent works well in this context:

1. Read and assimilate: studying course materials and theory provides the foundation.
2. Practice: chat with the agent and practice scenarios.
3. Apply: make decisions, solve problems.
4. Reflect and learn from experience: the agent provides feedback, and the students evaluates their learning.

Possible use cases 

The AI agent is suitable for almost any field that requires human interaction, problem-solving, or interpretation of situations. Here are some examples:

• Customer service: The agent can act as a customer in any field the student is studying. The customer asks questions or presents a problem, and the student practises service orientation, problem-solving, guidance, and resolving service requests from their own field.
• Leadership and management: Use cases include simulating discussions or  interviews, conflict resolution and mediation, board work, and encounters with investors or financiers.
• Project work: The agent can be the “project client” providing feedback or making decisions. The student practises communication, negotiation, and project management in practice, as well as presenting a new solution or project plan to the client.
• Technical fields: Use cases include problem and case exercises, software development with the client, eliciting customer requirements and cases, and scenarios where the student explains a technical solution to a “non-technical” agent.
• Nursing and medicine: The agent can be used for simulating patient encounters, communication exercises with patients of different ages or backgrounds, and practising challenging situations.
• Communication and interaction: The student can practice planning a campaign, announcement, or communication with a client, crisis communication cases, and receive feedback on scripts and texts.
• Education: The agent can guide planning processes and act as a customer. An exercise can also involve teaching a subject to a “student,” as well as advice and guidance discussions.
• Legal advice: The student can practice resolving situations and explaining solutions to a client. The agent can invent cases and act as a witness, opposing party representative, or judge.
• Languages: The student can practice the use of foreign language in various scenarios with the agent. The agent can give feedback on style, vocabulary, interaction, and cultural understanding.
• Culture and history: Use cases include historical context, source criticism, evaluating eras and cause and effect relationships.
• General Study Skills: Use cases include guidance in writing research plans and evaluating structure, practising writing and technical correction, justifying one’s own views and practising argumentation skills, debate, interview skills, group situation training, etc.

Instructions for teachers: How to use an AI agent in a course

Define the objective of the exercise:  What skill/content will the student practise with the agent? For example, customer service, project work, workplace interaction, advice, teaching a subject, explaining problem-solving… The agent can be given the same background material the student has studied, so it can test learning and give feedback.

Create the agent and clearly define its role: What role does the agent have? Customer, patient, colleague, project stakeholder, employer, layperson to whom the subject must be explained understandably. A precise role helps the agent act consistently and support the learning objective.When creating the agent, give it instructions for acting in the role. Specify, for example, that the agent does not provide ready-made answers and that its task is to spar, challenge, and ask follow-up questions. The agent must use course content and terminology. The agent should not correct the student too early but allow the student to try their own approach.

Copilot Chat is freely available to the staff and students of the Tampere higher education community at microsoft365.com/chat. Remember to sign in with your TUNI account; otherwise, you will use a less secure free version! Go to Copilot Chat, and you will see the Copilot Agents sidebar on the right. The list shows the agents you have already created and, below them, the Create Agent (New Agent) button.

Step-by-step instructions for creating a Copilot Chat agent can be found in this tip: Build AI Assistants with Copilot agent builder.

Example:  
In this case the student is intended to practise a customer scenario with the agent, acting as a planning officer for an educational organisation. In this role, the student meets the agent, who acts as the representative of a client organisation and engages in a discussion to determine the organisation’s needs for AI training. Based on the information received in the discussion, the student prepares a training plan and presents it to the client. The agent-client answers the student’s questions, asks clarifying questions, and evaluates the training plan. When the student ends the situation, they receive an assessment from the agent on how the customer situation progressed.

Remember, you can describe such a case to Copilot Chat and ask it to create a prompt, which you can enter as the agent’s definition.

The agent’s definition prompt in this case could be as follows:

Your role: Your task is to have a discussion with the student. In the discussion, you are the manager of an organisation in a client meeting. Your organisation wants to order a training plan from the student’s company. The student, with whom you are discussing, is a training planner working for a company that provides training services. Your organisation wants to order a comprehensive skills development plan and an internal training programme on the use of AI (especially in work, processes, and productivity). You are responsible for the training objectives, target groups, and benefits. Stay in this role throughout the discussion.

Your goal: In your discussion, the student aims to find out the following from you:

• • What kind of skills the organisation’s employees need for AI use.
  • What are the objectives of the training.
  • What skills, levels, and role-specific needs exist in the organisation.
  • What restrictions, wishes, and priorities the client has.
  • What risks, concerns, and success metrics the organisation has.

Simulate a client meeting and answer the student’s questions. Let the student lead the discussion.

Principles of operation:

• • Do not give information all at once. Only answer the student’s questions.
  • Your answers may be incomplete or require clarifying questions, as in typical client situations.
  • You may raise challenges, uncertainties, internal policies, and organisational wishes.
  • You can also ask your own questions to the student, for example: “How would you suggest we address this need?”, “Could you clarify what level would be suitable for basic users?”, “How would we measure the effectiveness of the training?”
  • Always remain in the client’s perspective: you do not know the details of training planning, but you know your organisation’s needs, challenges, and objectives well.
  • Do not provide ready-made training plans or structures – that is the student’s responsibility. You only react, comment, and ask for more information.

Behaviour:

• • Be realistic in the role of client: express needs, limitations, and expectations.
  • Your answers can be partial, demanding, interested, worried, or encouraging.
  • You may alter or specify your wishes as the discussion progresses, as happens in real client dialogues.
  • Do not leave your role at any point, except when the student asks for feedback after stepping out of role.

Structure of the discussion:

• • The student begins by interviewing you, the organisation’s manager, to find out the organisation’s needs.
  • You answer questions in role, not revealing all information at once.
  • You may add challenges and raise requirements (budget, schedule, role-specific needs, data protection, change support).
  • When the student presents their training plan: ask critical or clarifying questions. Challenge their reasoning. Highlight risks or additional wishes.
  • Your aim is to act as a sparring client, not an evaluator or teacher.

What must not be done:

• • Do not provide solutions or plans on the student’s behalf.
  • Do not give a ready-made training programme or module structure.
  • Do not offer theoretical information about AI, but remain in the client’s perspective.
  • Do not provide too much information unless the student asks for clarification.
  • Do not leave your role.
  • Do not suggest ready-made prompts.

Key mentality:

• • You are a knowledgeable but busy client who wants a good training programme.
  • You want to ensure that the student knows how to ask the right questions and build a workable plan.
  • Your task is to help the student practise real client work, not teach them content.

Main task throughout the exercise: Act as a sparring, realistic client, i.e., a company director who only provides information via the student’s questions, asks their own questions, and critically but constructively assesses the training plan presented by the student.

For the Suggested Prompts section, you can write a starting prompt that the student can click to get the conversation going in the right direction:

“Hello, glad we could arrange a meeting. What kind of training do you need for your organisation?”

You can inform students to stop the simulation with the following prompt:

“The client meeting has now ended: Please provide feedback on how I did in the meeting?”

Agent will provide an feedback about the meeting and suggestions how to improve discussion.

Test the agent thoroughly and adjust the prompt as needed. Share the link to the agent with students, for example, through your Moodle area.

Prepare instructions for students on using the agent

The student must understand that the agent is not an answer machine but a training partner. Provide the student with clear instructions on what advance preparation is expected. Describe the roles of the student and agent precisely and state the ultimate goal of the conversation, the phases involved, and how long the discussions should last. You cannot view the discussions that have taken place, so if you want to include the exercise in assessment, you must ask the student to save and report the discussion, and, for example, reflect on what they have learned in a learning diary. The student can continue the conversation with the agent from where they last left off and save the exercises. Conversations are found on the right side of the page under the Chats section. Students can use the agent independently, but the teacher’s job is to provide support if necessary. Also explain what to do if the agent does not work as expected.

Read more

All tips in the digital toolkit