BS TEACHING GUIDE

Course: Virtual Assistants with Generative AI

Duration: 3 hours (in-person)

Instructor: [Professor’s Name]

Program: Master in Digital Marketing & Artificial Intelligence

Tools: ChatGPT Plus (GPT-5) and Gemini Pro

🎯 Learning Objectives

• Understand the fundamentals of virtual assistants powered by generative AI.
• Identify the main differences between ChatGPT Plus and Gemini Pro in assistant creation.
• Design and build a functional assistant with a clear role, tone, and purpose.
• Apply prompt-engineering principles to improve assistant coherence and accuracy.
• Evaluate and compare the performance of two generative AI models through hands-on exercises.

🧠 Methodology

• Brief theoretical introduction supported by live demonstrations.
• Active learning approach: individual and small-group projects.
• Comparative experimentation: testing the same scenario across two AI models.
• Guided discussion on real-world applications and ethical implications.

📘 Session Structure

Module	Content	Duration	Methodology	Tools
1️⃣ Introduction & Context	Overview of Virtual Assistants and their evolution in the Generative AI era	20 min	Lecture + discussion	Slides
2️⃣ Fundamentals of Generative Assistants	LLMs, prompting, ChatGPT vs. Gemini, key use cases	30 min	Lecture + demo	ChatGPT & Gemini
3️⃣ Exercise 1 – Building a Basic Assistant	Define the assistant’s purpose, tone, and behavior	40 min	Guided workshop	ChatGPT Plus
4️⃣ Architecture & Training	Context, memory, conversational logic, and integrations	30 min	Lecture + mini-demo	ChatGPT Plus
5️⃣ Exercise 2 – Model Comparison	Apply same task in both ChatGPT and Gemini; analyze results	40 min	Practical group work	ChatGPT & Gemini
6️⃣ Wrap-Up & Discussion	Best practices, ethics, and applications in marketing and business	20 min	Open discussion	Classroom

🧩 Practical Activities

Activity	Description	Expected Outcome
Exercise 1	Build a fully functional assistant in ChatGPT Plus	Assistant with role, tone, and purpose clearly defined
Exercise 2	Run and compare both assistants in ChatGPT & Gemini	Analytical summary of model differences and strengths

📊 Evaluation Criteria

Criterion	Weight
Active participation	20%
Exercise 1 (Assistant creation)	40%
Exercise 2 (Model comparison)	30%
Final reflection and discussion contribution	10%

🧰 Required Materials

• Laptop with internet connection.
• Access to ChatGPT Plus (GPT-5) and Gemini Pro.
• Instructor’s slide deck (provided at session start).
• Comparison worksheet (distributed during the second half).

 

 

Module 1 — Introduction & Context

Duration: 20 minutes

Format: Lecture + brief discussion

Objective: Introduce the concept, evolution, and impact of virtual assistants powered by Generative AI.

Opening (2 min)

Good morning everyone, and welcome to our session on Virtual Assistants with Generative AI.

Today, we’ll explore one of the most transformative applications of artificial intelligence — conversational agents that don’t just respond, but truly understand, adapt, and create.

Over the next three hours, you’ll not only learn the theory behind these systems but also build and test your own virtual assistant using ChatGPT Plus and Gemini Pro.

This first part — our introduction — will set the stage. I’ll guide you through what virtual assistants are, how they’ve evolved, and why generative AI has completely changed the game.

Part 1 — What is a Virtual Assistant? (5 min)

Let’s start with a simple definition.

A virtual assistant is a software agent that can perform tasks or services based on user input — usually through natural language.

Early examples include Siri, Alexa, and Google Assistant. These were built primarily around predefined scripts and command recognition. They could tell you the weather, set reminders, or play music — but they couldn’t think beyond their training.

Today, we’re dealing with Generative Virtual Assistants — assistants powered by Large Language Models, or LLMs, like ChatGPT and Gemini.

These assistants don’t just follow a script. They generate responses dynamically, drawing on massive amounts of contextual knowledge, and adapting to the conversation in real time.

In other words: they’re not just reactive, they’re creative.

Ask the class: How many of you have already used ChatGPT or Gemini for a professional purpose — not just curiosity?
(Wait for a few hands) Good. That’s exactly the mindset we’ll build on today — moving from using them as tools to designing intelligent assistants that think strategically.

Part 2 — From Chatbots to Generative Agents (6 min)

Let’s take a quick look at the evolution of virtual assistants.

Phase 1 – Rule-based Chatbots (2000s): These were decision-tree systems. Every possible question had a predefined answer. Great for FAQs, terrible for human-like interaction.

Phase 2 – NLP-based Assistants (2010s): Systems started understanding intent. Siri, Alexa, and Google Assistant used statistical NLP — they could classify commands like ‘Call Mom’ or ‘Set an alarm’, but they still didn’t understand context.

Phase 3 – Generative AI (2020s): The revolution came with models like GPT, PaLM, and Gemini. Now assistants could generate new content, reason across domains, and adapt tone and style.

What changed wasn’t just the size of the model — it was the shift from retrieval to generation. Instead of choosing from a set of responses, the AI creates a new one every time.

This is what allows virtual assistants today to act as customer support agents, marketing strategists, personal tutors, and even creative collaborators.

Part 3 — Why Generative AI Changes Everything (5 min)

Let’s connect this to real business and marketing applications.

With Generative AI, assistants can now:
– Understand brand voice and tone.
– Adapt to customer emotion and intent.
– Personalize communication at scale.
– Integrate with data sources for real-time responses.

For example, imagine a fashion brand using a ChatGPT-based assistant to help customers choose products, or a university using a Gemini-powered agent to guide students through enrollment.

The assistant doesn’t just provide information — it represents the brand.

That’s why, as marketers and strategists, it’s crucial to understand how to design assistants that communicate intelligently, ethically, and effectively.

Part 4 — The Human + AI Partnership (2 min)

Here’s something important: these assistants are not replacing humans — they’re amplifying human capacity.

The most successful organizations use AI assistants to handle repetitive, data-driven tasks, freeing people to focus on creativity, empathy, and strategy.

The real power lies in human–AI collaboration — when the assistant becomes an extension of the professional’s knowledge and personality.

Discussion Prompt (Final minute)

Before we move to the next section, I’d like you to think about this:

If you could design your own virtual assistant for your work today, what would it do for you?

We’ll return to that idea later when you start building your own assistant in ChatGPT.

Summary (30-second wrap-up)

So, to summarize:
– Virtual assistants have evolved from rule-based bots to generative partners.
– Generative AI allows them to reason, adapt, and create.
– The key opportunity is designing assistants that serve real human and business needs.

In the next module, we’ll explore the core mechanics of generative assistants — how LLMs work, what prompt engineering really means, and how you can control these models to produce exactly the responses you need.

 

 

Module 2 — Fundamentals of Generative Assistants

Duration: 30 minutes

Format: Theoretical explanation + live demo

Objective: Understand how generative assistants work, what powers them, and how marketers can control their behavior through prompt design.

Opening (2 min)

Now that we’ve understood what virtual assistants are and why they matter, let’s dive into how they actually work.

This module is about the foundation — the technology, logic, and structure behind these assistants.

But don’t worry — I’m not going to turn you into computer scientists. My goal is for you, as marketers, to think strategically about how these assistants can be designed, guided, and optimized.

Part 1 — How Generative AI Works (8 min)

At the heart of every modern assistant lies a Large Language Model — what we call an LLM.

You can think of an LLM as a statistical engine trained on enormous amounts of text — billions of words from books, articles, and the web. It learns patterns: how humans use language, structure sentences, and convey meaning.

When you ask a question, it doesn’t look up an answer — it predicts the most likely next words, based on everything it has learned.

Let’s simplify:
– Traditional software follows rules.
– Generative AI learns from examples.

This difference is what gives tools like ChatGPT or Gemini their creative ability — they generate content that has never been written before.

Example:
If I ask a rule-based chatbot: ‘Create a social media caption for a coffee brand,’ it would look for a stored template.

A generative model, however, will produce a new caption each time — one that matches tone, emotion, and brand context. That’s creativity through prediction.

Part 2 — The Role of Prompting (8 min)

Now let’s talk about your main superpower as marketers working with these systems — prompting.

A prompt is simply the instruction you give the model. But the difference between an average and a brilliant assistant often comes down to how you write that prompt.

Good prompts provide:
– Context (Who is speaking? What’s the situation?)
– Role (What is the assistant supposed to be?)
– Goal (What do you want as output?)
– Tone (Professional, persuasive, friendly, analytical, etc.)
– Format (Table, email, ad copy, strategy, etc.)

In marketing, this means you can shape the assistant’s behavior to act like a copywriter, strategist, or customer service rep.

Example demonstration in ChatGPT Plus:
Prompt: ‘Act as a senior marketing strategist for a luxury fashion brand. Write a 3-line Instagram caption promoting a new limited-edition handbag in a sophisticated tone.’

Notice how the assistant instantly adopts a persona, tone, and purpose — all through the prompt.

That’s the magic of prompt engineering — it’s communication design for AI.

Part 3 — ChatGPT Plus vs Gemini Pro (7 min)

Now that we know the logic, let’s compare the two main tools we’ll use: ChatGPT Plus and Gemini Pro.

Both are generative AI systems, but they have slightly different architectures and personalities.

ChatGPT Plus (GPT-5)	Gemini Pro
Strength: Language fluency, creativity, reasoning	Strength: Multimodal understanding (text + images + data)
Style: Conversational, adaptable, strong structure	Style: Analytical, fact-based, efficient
Interface: Custom GPTs and assistant design flexibility	Interface: Integrated within Google ecosystem
Ideal for: Marketing copy, storytelling, assistant design	Ideal for: Market research, factual analysis, structured tasks

In practice, ChatGPT is great for creativity, storytelling, and tone design — while Gemini excels in analysis, data grounding, and structured tasks.

Later in the course, you’ll test both using the same prompt and see how their reasoning differs.

Part 4 — Anatomy of a Generative Assistant (4 min)

Let’s break down the anatomy of a virtual assistant you might create:
1. Purpose: What problem does it solve? (e.g., helping users plan campaigns)
2. Personality: Formal, playful, expert, empathetic… tone matters.
3. Knowledge Base: Does it use uploaded documents or contextual data?
4. Interaction Flow: How does the user talk to it — open chat, menu, or guided flow?
5. Limitations: What should it not do or say?

Designing an assistant is like brand building — defining identity, boundaries, and voice.

Reflection:
Think of your favorite brand. If it had a virtual assistant, how would it sound? What would it avoid saying? This helps you translate brand identity into conversational behavior.

Part 5 — Good vs Bad Assistants (3 min)

Let’s clarify what makes a good virtual assistant in marketing.

Bad assistants:
– Sound robotic or generic.
– Give inconsistent or off-brand answers.
– Fail to adapt to the user’s tone.

Good assistants:
– Use natural, human-like communication.
– Maintain tone and empathy.
– Provide value quickly and coherently.

As marketers, you must ensure your assistant delivers the brand experience — not just the right words.

Discussion (2 min)

Before we close this module:
Think about one specific area in digital marketing — content creation, SEO, social media, email, paid media…
Where do you believe generative assistants could bring the biggest transformation?

We’ll share a few ideas before we move to the practical module.

Summary (1 min)

To summarize:
– Generative assistants are powered by LLMs that learn from massive text data.
– Prompts are your design tool — they control behavior, style, and outcome.
– ChatGPT and Gemini have complementary strengths — creativity vs data grounding.
– Designing an assistant requires defining purpose, personality, and boundaries.

Next, we move from theory to practice — in the next module, you’ll create your first fully functional assistant.

 

 

Module 3 — Building a Basic Assistant

Duration: 40 minutes

Format: Guided workshop (hands-on in ChatGPT Plus)

Objective: Design and build a functional virtual assistant that has a clear purpose, role, tone, and personality — aligned with marketing objectives.

Opening (3 min)

Now we move from theory to practice.

You’ve learned what generative assistants are and how they work. In this module, you’ll build your first assistant — step by step — using ChatGPT Plus.

Think of this as your first digital employee: it will have a name, a purpose, and a personality. Our goal today is to make it useful, human, and aligned with a brand voice.

Part 1 — Defining the Purpose (7 min)

Every assistant starts with a clear ‘why’.

Ask yourself three key questions:
1. What problem does this assistant solve?
2. Who will use it?
3. What is the main action or value it delivers?

Examples:
– A customer-service assistant for an e-commerce brand.
– A content-generation assistant for a social-media agency.
– A data-insight assistant for a marketing analytics team.

Activity: Write your assistant’s purpose in one sentence:
‘This assistant helps [target user] to [achieve a goal] by [doing what].’

Part 2 — Designing the Role and Personality (7 min)

Now that we know the purpose, we give the assistant a role and voice.

Every assistant has a tone and a personality. Think of:
– Formality: formal / friendly / conversational.
– Tone: analytical / optimistic / empathetic / humorous.
– Expertise: beginner / professional / senior strategist.

Activity: Describe your assistant’s personality in 3 adjectives (e.g., Empathetic, strategic, inspiring).

Part 3 — Writing the Prompt (12 min)

Now we turn your idea into an operational assistant.

In ChatGPT Plus → Explore GPTs → Create a new GPT.

Example prompt:
‘Act as a professional social media strategist for a sustainable fashion brand. Your mission is to help the marketing team create engaging posts, generate captions, and suggest hashtags consistent with the brand’s tone: elegant, inspiring, and eco-conscious. Always write in British English and limit captions to 30 words.’

A good assistant prompt includes:
1. Role definition
2. Objective
3. Brand tone
4. Constraints
5. Output examples.

Mini-exercise: Write your own prompt, paste it into ‘Instructions for the assistant’, and give your assistant a name.

Part 4 — Testing and Refining (8 min)

Now test your assistant with 3 queries:
1. Simple: ‘Write a 3-line caption about vegan sneakers.’
2. Complex: ‘Create a 1-week posting plan with channel and objective.’
3. Boundary: ‘Write something funny but still elegant.’

Evaluate if it follows tone, brand, and precision.

If not, refine prompts — add tone or word limits.
Assistant design is iterative — like campaign optimization.

Part 5 — Class Reflection and Discussion (5 min)

Let’s share a few examples.

Who built a marketing assistant? A customer-service one? A research one?
What tone worked best? Which answers surprised you?

The power isn’t in using ChatGPT — it’s in guiding it strategically.

Summary (3 min)

You’ve defined your assistant’s purpose, tone, and prompt.
You’ve built and tested a real assistant inside ChatGPT Plus.

Next: We’ll explore training assistants with custom data to make them smarter marketing partners.

 

 

Module 4 — Architecture & Training of Generative Assistants

Duration: 30 minutes

Format: Lecture + mini-demo

Objective: Understand how to structure, contextualize, and train assistants to perform complex, brand-aligned tasks using external knowledge.

Opening (3 min)

Until now, you’ve created an assistant that can speak and act — but its knowledge is limited to what the model already knows.

In this module, we’ll learn how to make assistants smarter and specialized. We’ll see how to give them memory, context, and access to custom information — turning them into true digital experts for marketing or communication tasks.

Think of this as turning your assistant from a generalist into a specialist.

Part 1 — Core Architecture of a Generative Assistant (7 min)

A Generative Assistant typically has four layers:
1️⃣ The Model (LLM): the brain — interprets and generates language.
2️⃣ The Instructions / Prompt Layer: personality and rules you define.
3️⃣ The Memory or Knowledge Layer: contextual information or uploaded data.
4️⃣ The Interface Layer: how users interact — chat, app, or API.

In marketing terms:
• The LLM is your strategist.
• The prompt is your brand manual.
• The knowledge base is your marketing database.
• The interface is your campaign touchpoint.

Part 2 — How Training Works (7 min)

Training assistants doesn’t mean retraining the model — that’s done by OpenAI or Google.

We use contextual training to enhance reasoning:
🔹 Prompt-based training — refining with examples and tone.
🔹 File-based training — uploading documents, FAQs, or reports.
🔹 Memory or fine-tuning — remembering recurring data or patterns.

Example: Uploading property data, market reports, and tone guides for a real estate assistant ensures brand-aligned outputs.

Part 3 — Designing a Knowledge Strategy (6 min)

Plan the data you feed your assistant carefully:
1️⃣ Relevance — include only useful data.
2️⃣ Quality — ensure accuracy and updates.
3️⃣ Structure — use clean, labeled formats.
4️⃣ Boundaries — avoid confidential content.

Example (retail brand):
• Upload product catalog.
• Upload brand tone guidelines.
• Upload customer FAQs.
• Add sample campaigns.

This combination ensures depth and consistency.

Part 4 — Integrations and Automations (5 min)

Once structured and trained, assistants can integrate into workflows:
• CRM integration — personalize via HubSpot or Salesforce.
• Social media — connect to Meta or X for content publishing.
• Email automation — connect to Brevo or Mailchimp.
• Data dashboards — connect to Looker Studio.

This creates an ‘AI in the loop’ system — AI sits between humans and data to enhance marketing performance.

Part 5 — Ethical & Practical Considerations (4 min)

Guidelines for marketers:
• Transparency — disclose AI interaction.
• Accuracy — verify all data sources.
• Privacy — protect sensitive information.
• Bias — ensure inclusivity and fairness.

Assistants represent not just your brand technically, but ethically.

Summary (3 min)

• Generative assistants = model + instructions + knowledge + interface.
  • Train with contextual data, not by changing model weights.
  • Marketing assistants thrive on structured, verified information.
  • Integrations turn assistants into workflow partners.
  • Ethics and transparency build trust.

  Next: Exercise 2 — training and comparison with Gemini Pro.

 

 

Module 5 — Exercise 2: Training and Model Comparison

Duration: 40 minutes

Format: Practical workshop (hands-on comparison)

Objective: Train assistants with real contextual data and evaluate qualitative and functional differences between ChatGPT Plus (GPT-5) and Gemini Pro.

Opening (3 min)

Now it’s time to put everything together.

You’ve already designed your assistant — it has a personality, tone, and structure. In this module, you’ll make it smarter by adding contextual data and then compare ChatGPT Plus and Gemini Pro.

This exercise lets you see how different models think — and helps you pick the right one for each marketing task.

Part 1 — Setting Up the Context (5 min)

Before training, define what knowledge your assistant needs.

Choose one of the following cases:
• Case A: Sustainable fashion brand assistant.
• Case B: E-commerce customer service assistant.
• Case C: Content planner for a tech startup.

Write a short brand brief (100–150 words) describing tone, values, and audience. This becomes your training document.

Part 2 — Training the Assistant (10 min)

Upload or paste your brief into ChatGPT Plus under ‘Knowledge’ or the main prompt.

Example:
‘Use the following brand brief to guide all your answers: [insert text].’

Then test it:
1️⃣ Create a product description for your brand.
2️⃣ Write a short promotional email.
3️⃣ Generate a 1-week content plan for Instagram.

Observe tone, style, and creativity.

Part 3 — Testing Gemini Pro (10 min)

Now repeat the same steps in Gemini Pro with the same brief and prompts.

Gemini is more analytical; ChatGPT is more creative. Compare both:
• Product description
• Email campaign
• Content plan

Take notes on structure, depth, and tone differences.

Optional demo: instructor shows both model responses side by side.

Part 4 — Comparison & Evaluation (8 min)

Evaluate each model using these criteria (1–5 scale):

| Criterion | Description |
|————|————-|
| Creativity | Originality and expressiveness |
| Accuracy | Brand and factual consistency |
| Tone Adaptation | Alignment with audience |
| Clarity & Usability | Ease of marketing implementation |

Discuss results: Which model better fits your brand voice?

Part 5 — Class Discussion (3 min)

Share your findings:
• Which model felt more natural?
• Which followed tone better?
• Which output surprised you?

Understanding these differences helps select the right tool for ideation, research, or execution.

Summary (1 min)

You have:
• Trained your assistant with contextual data.
• Compared ChatGPT Plus and Gemini Pro.
• Evaluated creativity, tone, and accuracy.

Next: Module 6 — conclusions, best practices, and ethical insights.

 

 

Module 6 — Conclusions and Best Practices

Duration: 20 minutes

Format: Discussion + reflection

Objective: Summarize the main learnings, highlight best practices, address ethical implications, and outline real-world applications of generative assistants in marketing and communication.

Opening (2 min)

We’ve reached the final