Classic Bot vs Modern Copilot
Continuing the powerful capabilities of Microsoft Copilot Studio, here comes two of the most powerful approaches to making your chatbot framework, i.e. Classic Bot and Modern Copilots. Both these have minor differences and great use according to your framework.
In this tutorial chapter, we will see the distinctions between these two types of bots, how they work, the extensions they support and a detailed comparison of their capabilities. By the end, you will have a clear idea to decide the perfect approach for your project.
What is a Classic Bot in Copilot Studio?
A Classic Bot in Microsoft Copilot Studio is built around rule-based, predefined logic. These bots operate on decision trees, using fixed paths to guide conversations. Classic bots are reliable for simple, repetitive tasks like handling FAQs, basic customer support, or preprogrammed interactions.
Steps for Creating a Classic Bot
- Open Microsoft Copilot Studio and click on the Create Classic Bot option.
- Select from available Prebuilt Templates like Customer Support or FAQ Bot.
- Define Intents and Utterances −
- Intents are actions the bot needs to respond to, such as "Place an Order" or "Get Weather Info".
- Utterances are examples of what users might say to trigger those intentions.
- Set Response Patterns for each intent: Fixed responses based on decision trees.
- Test the bot in the built-in test environment by providing different user inputs and verify the responses match the programmed rules.
Example
If a user asks What is my account balance?, a Classic Bot can return a predefined response, such as Your balance is $1000, based on a fixed rule.
What is Modern Copilot in Copilot Studio?
Unlike Classic Bots, the Modern Copilot takes a machine-learning-driven approach to conversations. Modern Copilots are highly trained on the dataset available to Microsoft with its powerful ML models and AI algorithms which support real-time learning, and context-driven adaptability. They are powered by Microsoft Graph, allowing deep integration with user data and services.
Steps for Creating a Modern Copilot
- Open Microsoft Copilot Studio and navigate to the Create Modern Copilot section.
- Choose the Custom Entities and Machine Learning Models for Contextual Conversations.
- Define flexible conversation paths by allowing the Copilot to dynamically interpret user inputs using natural language processing (NLP).
- Add Context Management for personalised responses based on user data.
- Test the Modern Copilot using the live testing environment with a wide range of inputs to see how the AI adapts and evolves.
- Monitor and tweak responses using AI feedback loops to improve accuracy over time.
Example
When asked "Whats my next meeting?", a Modern Copilot can pull data from the users calendar using Microsoft Graph, offering real-time, personalised responses.
A Modern Copilot can handle a vague request like "Schedule my weekly report", interpret the context, and respond with personalised details based on the user's past interactions or calendar.
One of the main differences between a Copilot and a Classic Bot is the Generative AI feature, which is only available in Modern Copilot.
Real-Life Scenario Based on Classic Bot and Modern Copilot
1. Classic Bot in Action
- User − Sarah, an online banking customer.
- Scenario − Sarah needs to check her account balance and wants to find out the nearest branch of her bank for an in-person visit. She uses her banks customer service chat feature, powered by a Classic Bot.
- Initiation − Sarah opens the banks website and clicks on the chat icon.
User Input − Sarah types, "I need to check my balance".
Classic Bot Response − The Classic Bot has a predefined rule for this query. It checks for specific keywords like balance and responds with −
- "Please enter your account number or log in to your account to view your balance".
User Verification − Sarah enters her account number.
Classic Bot Response − The bot provides a static, predefined response −
- "Your current balance is $5000".
Additional Request − Sarah then asks, "Where is the nearest branch?"
Classic Bot Decision Tree − The bot follows its programmed path and asks Sarah for her zip code. Sarah replies with her zip code.
- Based on the input, the bot fetches a predefined list of branch locations and displays: The nearest branch is at 1234 Main Street, and it is open from 9 AM to 5 PM.
Hence, we can say that the Classic Bot is efficient in handling repetitive, straightforward queries like checking account balances or providing branch details. It operates within a fixed, rule-based framework, making it reliable but limited to predefined interactions.
Sarahs interaction could have been smoother if she needed more dynamic help. For example, if she wanted personalised advice on her account activity or assistance in scheduling a meeting, the bot wouldnt have been able to handle it efficiently without custom rules added for every possible interaction.
2. Modern Copilot in Action
- User − Sam, an employee working remotely.
- Scenario − Sam is working from home and needs to schedule a meeting with his team, pull relevant files from his companys SharePoint, and receive a summary of the last meetings notes. He uses a Modern Copilot integrated into his companys workflow system.
- Initiation − Sam opens his companys internal collaboration app that has the Modern Copilot embedded.
User Input − Sam types, "Schedule a meeting with my team for tomorrow at 10 AM".
Copilot Response − The Modern Copilot, using its natural language processing (NLP) capabilities, interprets Sams request. It connects to Sams calendar via Microsoft Graph and sends out meeting invitations to the team.
- "Your meeting with the team has been scheduled for tomorrow at 10 AM".
User Input − Sam follows up with, Can you pull up the latest project files from SharePoint?
Copilot Action − The Copilot dynamically connects to SharePoint, searches for files related to the ongoing project, and retrieves them in real-time.
- Here are the latest project files from SharePoint: [File1.pdf], [File2.docx]
User Input − Sam then asks, "Give me a summary of the last meetings notes".
Copilot Response − The Modern Copilot pulls the relevant meeting notes from OneNote, processes them using its AI capabilities, and returns a concise summary.
- "Here's a summary of the last meeting: Project Alpha is on track, with a deadline of next Friday".
After completing these tasks, the Copilot updates its knowledge base, learning from Sams preferences and context to offer even more efficient responses in future interactions. Hence, it dynamically processes information, adapting to Sams needs in real-time without requiring predefined rules for each specific action.
Key Differences Between Classic Bot and Modern Copilot
The following table highlights major differences between the Classic Bot and Model Copilot −
| Aspect | Classic Bot | Modern Copilot |
|---|---|---|
| Technology | Rule-based, decision-tree-driven | AI-driven, powered by machine learning and NLP |
| Based On | Power Virtual Agents | New Copilot functionality |
| Generative AI | - | Yes |
| Interface | Old | Modern |
| AI Capabilities | Very Limited | Fully Integrated with AI |
| Solution Management | Uses Topic component | Uses Topic V2 component |
| Plugin Support | - | Yes |
| Interaction Style | Limited to predefined responses and rigid workflows | Dynamic, adaptive, and context-aware interactions |
| Learning Ability | No learning capability; operates solely on predefined rules | Continuous learning through adaptive AI and user behaviour |
| Integration with External Systems | Limited integration, usually requires custom connectors | Deep integration with platforms like Microsoft Graph, SharePoint, OneDrive, etc. |
| Response Flexibility | Fixed responses for specific queries | Real-time, context-sensitive responses |
| Natural Language Understanding (NLU) | Minimal or none | Advanced natural language processing (NLP) |
| Handling Complex Queries | Poor, requires manual escalation | Excellent, can process and resolve complex multi-step requests |
| Context Awareness | Lacks the ability to maintain conversation context | Maintains and builds context over the course of interaction |
| Personalisation | No personalisation; same responses for all users | Tailored responses based on user preferences and history |
| Task Automation | Simple task automation with limited functionality | Advanced task automation with dynamic workflows |
| Scalability | Requires manual rule updates to scale | Scales automatically through AI and ML without manual intervention |
| Data Processing | Minimal; handles only predefined inputs | Real-time data processing from integrated systems |
| Training & Maintenance | Requires continuous manual updates | Self-learning; minimal manual maintenance |
| Error Handling | Follows rigid error paths; often leads to dead ends | Adaptive; offers alternative solutions and recommendations |
| User Experience | Static, repetitive interactions | Fluid, engaging, and personalised experience |
| Context Retention Across Sessions | No session memory; each conversation starts afresh | Retains user context and preferences across sessions |
| Deployment Flexibility | Requires custom deployment on specific platforms | Can be deployed across various environments with ease |
| Migration | Possible | N/A |
| Topics | Manual topic creation | Topic created through Copilots Generative AI |
| Triggers | Just trigger phrases | Supports multiple trigger phrases |
| Event | - | Send event, Send activity, Send HTTP request, Log custom telemetry event. |
In conclusion, both the Classic Bots and Modern Copilots have their own strengths and use cases. If you're looking for simplicity and predictability, Classic Bots may be the way to go. However, if adaptability, real-time learning, and integration with modern data sources are what you need, Modern Copilot is your best bet. Both approaches allow you to build robust conversational agents in Microsoft Copilot Studio, but the choice ultimately depends on your project's complexity and requirements.