# Lesson 6.2 – Training your Team on AI

{% embed url="<https://youtu.be/oNAl4kFDcHI>" %}

### 🎯 Learning Objectives

By the end of this lesson, your team will be able to:

* Understand how **prompting impacts AI output**
* Learn the basics of **effective prompt design**
* Use structured language to improve Agent accuracy and usefulness
* Apply prompting in **real-world daily workflows**
* Build confidence and habits around **AI interaction as a skill**

***

### 📌 Why Prompting Matters More Than You Think

<figure><img src="https://3805827895-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FSfECtcNwrIDQm7NrCIeB%2Fuploads%2FXQqpBnQRQRWpC1QL0OAQ%2Fimage.png?alt=media&#x26;token=958058c1-1c3e-4a20-95f1-e808ad8fa7f0" alt=""><figcaption></figcaption></figure>

AI Agents are not all-knowing—they’re context-sensitive tools. What you *ask* shapes what you *get*.

Unlike traditional software that follows buttons and rules, **AI follows language**.

A slight change in phrasing can:

* Improve clarity
* Reduce hallucinations
* Surface more relevant data
* Encourage longer or more concise answers

✅ Teaching your team to **speak AI** is one of the most important steps in successful adoption.

***

### 🧠 The Science of Prompting

<figure><img src="https://3805827895-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FSfECtcNwrIDQm7NrCIeB%2Fuploads%2FjtEn5CCvugyWXGVw4sfN%2Fultra_compact_prompting_guide.png?alt=media&#x26;token=7f761d6a-c199-4e35-a0f0-7c09954f05b0" alt=""><figcaption></figcaption></figure>

Every AI interaction starts with a prompt—a question, instruction, or command.

The better your prompt, the better the result.

**Prompting affects:**

* **What sources** the AI retrieves from
* **How much detail** it includes
* **Whether it stays grounded** in known information
* **Whether it needs to call a function or workflow**

***

### ✍️ Prompting Basics for Your Team

Here’s a simple framework for building better prompts:

| **Bad Prompt**    | **Improved Prompt**                                                                                |
| ----------------- | -------------------------------------------------------------------------------------------------- |
| “Refunds?”        | “What is our refund policy for digital products purchased more than 30 days ago?”                  |
| “Help onboarding” | “Can you walk me through the onboarding process for new remote employees in the sales department?” |
| “Send invoice”    | “Can you send the last invoice for account ID 234512?”                                             |

***

### 💡 Prompting Tips to Teach Your Team

| ✅ Do                                  | ❌ Don’t                           |
| ------------------------------------- | --------------------------------- |
| Be specific                           | Be vague or overly short          |
| Provide context                       | Ask out of the blue               |
| Clarify the format you want           | Assume the AI knows your intent   |
| Use natural but structured language   | Use slang or disconnected phrases |
| Break down complex asks into steps    | Ask for 10 things at once         |
| Say what *not* to include (if needed) | Leave scope open-ended            |

***

### 🧪 Prompting Is a Skill—Practice Makes Proficient

<figure><img src="https://3805827895-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FSfECtcNwrIDQm7NrCIeB%2Fuploads%2F0N4EjutDMtgB1J6gLR1W%2Fimage.png?alt=media&#x26;token=80341c90-81c9-431f-8aea-ab98e1086aac" alt=""><figcaption></figcaption></figure>

Just like writing good emails or giving good presentations, prompting gets better with:

* Repetition
* Review
* Feedback

Encourage your team to **experiment**, **iterate**, and **review results**. Over time, they’ll develop intuition for what works best.

***

#### 🔁 Prompting + Feedback = Continuous Improvement

<figure><img src="https://3805827895-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FSfECtcNwrIDQm7NrCIeB%2Fuploads%2FyxkgqrdrqVzzi2AieZrg%2Fimage.png?alt=media&#x26;token=c1ea9c7e-163c-45ab-90c6-23da97a7f428" alt=""><figcaption></figcaption></figure>

Good prompting doesn't just improve the answer—it improves the **Agent** over time.

Every time your team:

1. Prompts the AI
2. Rates the response as GOOD/BAD
3. Suggests a better version

→ The system **learns**. Your AI Agent evolves.

This turns your entire team into **collaborators in training the Agent**, not just users.

📘 Reinforced in \[Lesson 5.2 – Human Feedback with raia Copilot]

***

### 🛠 Training Employees to Use AI in Their Workflow

<figure><img src="https://3805827895-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FSfECtcNwrIDQm7NrCIeB%2Fuploads%2FbwFUjyL7nKFugqwdteai%2Fai_training_infographic.webp?alt=media&#x26;token=75158a5c-bf34-4d13-96cb-20e301ba2b58" alt=""><figcaption></figcaption></figure>

Don’t just tell employees *how* to use AI—show them *where* to use it.

Common use cases:

* Drafting responses (support, email, follow-up)
* Searching policy or procedural knowledge
* Creating summaries or reports
* Asking for next steps based on specific roles
* Triggering automated tasks via functions or workflows

**Encourage team members to:**

* Use Copilot or embedded chat regularly
* Log prompts that worked well
* Share “favorite prompts” or time-saving examples
* Track what areas the Agent is helpful (or not yet helpful) in

***

### 🧪 Real-World Prompting = Real Training

Every time someone uses the AI Agent to:

* Solve a real problem
* Ask a recurring question
* Test a live task

→ They’re helping **train it better**.

This is how AI becomes **embedded** in operations—not just tested in isolation.

***

### 📝 Prompt Training Exercise

<figure><img src="https://3805827895-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FSfECtcNwrIDQm7NrCIeB%2Fuploads%2FWF785PFuHFqw8W55nUh2%2Fimage.png?alt=media&#x26;token=b042a26d-4f7a-42a4-82b1-a7092695a590" alt=""><figcaption></figcaption></figure>

Have team members try the following:

1. Think of a task they do regularly (e.g., “Answering product questions”)
2. Write 3 prompts for the Agent to help with that task
3. Evaluate the responses
4. Refine and retry using what they learned
5. Submit one “most effective” prompt to the team wiki

***

### 👥 Prompting Training Do’s & Dont's (Summary Card)

| ✅ DO                            | ❌ DON’T                              |
| ------------------------------- | ------------------------------------ |
| Be specific and structured      | Be vague                             |
| State what you want in output   | Leave it open-ended                  |
| Reference roles or personas     | Ask abstractly                       |
| Use “step-by-step” or “show me” | Assume the Agent knows your thinking |
| Try multiple versions           | Blame the Agent on first try         |

***

### ✅ Key Takeaways

* Prompting is a **critical skill** that drives AI quality
* The more specific and structured the prompt, the better the answer
* Encourage real-world use to train the Agent *and* improve workflows
* Build prompt fluency across the team to maximize value
* AI is a collaborator—treat interactions as training, not transactions