Chatbot Showdown: Microsoft Copilot Studio vs. Google Dialogflow CX

Microsoft Copilot Studio and Google Dialogflow CX are both chatbot creation tools. We’ll start with Dialogflow as the baseline and compare its capabilities to the competitor’s. If there are any unique or impressive features of the competitor, we’ll highlight them.

Executive Summary:

One significant difference will be evident even before we delve deeper into the comparison. Microsoft Copilot Studio is marketed as a no-code product that operates within the boundaries of the provider’s platform. It’s primarily designed for non-technical users who focus on the business aspects of chatbot development. The technical aspects of the product are intended to be limited and controlled by the owner.

In contrast, Dialogflow offers flexible and complete API control over chatbot conversations. This allows almost every aspect of the conversation to be customized, enabling users to fully harness the power of modern programming tools.

סיכום: שתי הפלטפורמות נראות יכולות להביא פתרון יציב לצ׳ט המבוסס על מידע עירוני. יחד עם זה מערכת שתושתת על קופיילוט יכולה לקום הרבה יותר מהר על חשבון יכולות מתקדמות של הצ׳ט ועקומת הלימוד יותר נמוכה. קופיילוט לא מצריך ממפתחים יכולות קידוד ושימוש בכלים הלא מוכרים בארגון. הפריסה של המערכת ועיקר התמיכה שלה מתבצע בסביבה הענן המוכרת ומתוחזקת היטב ארגון - MS Azure. קופיילוט היא פלטפורמה חדשה יחסית וחוץ מאנשי Microsoft, עוד לא נוצרה קהילה רחבה של המשתמשים בה שיכולים לחלוק את ניסיונם מהיבטים שונים.

להבדיל מזה, הפלטפורמה Google Dialogflow CX יכולה לתת מענה לתסריטי שיחות מורכבים יותר ויחד עם זה להיות בבסיס ליכולות חיפוש ושיוך מסמכים בעירייה. הבחירה בה נראית כרוכה ביכולות פיתוח גבוהות של המפתחים ושילובים במסגרות לא מוכרות בעיריית ת״א. התמיכה מצד של Google גם לא נראית מקיפה ולרוב שטחית ולוקחת זמן רב. מצד שני, זאת פלטפורמה בת כשמונה שנים וקיים חומר גדול ברשת אודות שימוש בה מהקהילה.

עלויות של תחזוקה של שתי המערות דומות, אבל תחזוקת עצמה של קופיילוט נראית טבעית יותר לארגון עם ניסיון רב במערכות Microsoft.

High-level Feature Comparison

Feature	Dialogflow CX	Copilot	Win/Lost
Hebrew Language Support	Full Support, except RTL in DF-Messenger	Experimental, but seems mostly completed	Mostly the same
Easy of use	Not easy, steep learning curve.	Much more friendly, low learning curve.	Copilot win
Advanced Features	Support  for chat sessions/history. NLU for classification, partial responses, statistical analysis via built-in functions.	No chat session restoring, ability to execute custom actions, including Power Automate flow. Agent’s triggers.	Dialogflow win
Graphical design	Built-in UI designer for chat flows. Involves programming for return cards.	No programming for flow design and return cards. Built-in UI designer for chat flows. Adaptive Cards designer. This feature saves a lot of development time and support.	Copilot win
Publishing	Built-in into HTML pages, integration with Twilio, Facebook, etc.	Mostly the same	Equal capacities
API	Dedicated API (REST & gRPC)	Direct Line for conversation operations, Bot Framework REST API for bot creation.
Debugging	Built-in tester. External tools for advanced scenarios.	Build-in tester, common debugging as for Power Apps.	Copilot wins
Price	See here	See here
Datastore/Knowledge source	Automatic embedding for HTML pages, PDFs, FAQs docs.	Half-Automatic embedding for HTML pages, Sharepoint sites, dataverse (tables)	Copilot wins because built-in Sharepoint support.
Search capabilities	Extendable	?
Voice support	Yes	Yes	Equal capacities
Pictures/Video support	No built-in support, could be extended thru external AI services.	?
Supported LLMs	Built-in Gemini LLM, other models are available through the hooks.	Built-in ChatGPT, other models are available through Power Apps HTTP calls.	Mostly the same

Detailed comparison

Creating a chatbot involves designing the logic behind user-bot conversations in natural language. Both products support Hebrew, with Dialogflow providing full support and Copilot offering experimental support.

Dialogflow models the conversational domain as a set of “flows” consisting of “pages” that correspond to the states in a state machine. When a specific page in the flow is reached, the system responds with a “fulfillment” that is highly customizable.

The flow transitions from one page to another using “routes” that are activated based on the intent detection process. The user’s utterance is associated with the previously formulated intent as a set of phrases. This detection process is governed by Google NLP, specifically BERT and LaMDA.

In Copilot, the concept of “flow” is analogous to “topic.”

The term “intent” is absent from Copilot because the conversation domain modeled there differs. It primarily focuses on business logic rather than deep NLP customization.

Flow creation is a visual process in both products. For Microsoft users, the knowledge of Power Platform is quite comfortable. In fact, building a topic in Microsoft Copilot is essentially the same as building a Power Automate app in Microsoft Azure. In both cases, the assumption of no-code is strong.

Customization and Extensibility.

As is well-known, Power Apps does not allow calling external services directly, but rather through Power Automation’s flow. In contrast, DF widely uses HTTP REST callbacks to provide any necessary data on behalf of the conversation.

Publishing a chatbot.

This is a process of integrating external tools to interact with the designed logic.

The data protocol between the DF engine and the UI is well-defined and implemented in the HTML custom component DF-Messenger, which can serve as a reference implementation. However, this capability is not present in Microsoft Copilot Studio. The designed agent can only be published into pre-existing channels. Some customization can be achieved using Direct Line Secret and the Bot Framework, which provides a web chat script that can be added to HTML.

For example:

<!DOCTYPE html>
<html lang="en">
<head>
    <meta charset="UTF-8">
    <meta name="viewport" content="width=device-width, initial-scale=1.0">
    <title>Copilot Studio Chatbot</title>
    <script src="https://cdn.botframework.com/botframework-webchat/latest/webchat.js"></script>
</head>
<body>
    <h2>Copilot Studio Chatbot</h2>
    <div id="webchat" style="height: 500px; width: 400px; border: 1px solid #ccc;"></div>

    <script>
        (async function() {
            const directLine = window.WebChat.createDirectLine({
                token: "YOUR_DIRECT_LINE_SECRET"  // Replace with your Direct Line token
            });

            window.WebChat.renderWebChat(
                { directLine, userID: "user-123", username: "User" },
                document.getElementById("webchat")
            );
        })();
    </script>
</body>
</html>

Const directLine = window.WebChat.createDirectLine({**Direct Line** is a Microsoft service that allows you to build chatbots and integrate them with your applications. It provides a range of features, including direct line tokens, user IDs, usernames, and the ability to render WebChat.

Copilot, on the other hand, is an AI-powered tool that integrates seamlessly with OpenAI’s LLM. It offers a range of features, including RAG-based capabilities, the ability to use custom sites or SharePoint for RAG, and the option to use alternative generators via hooks.

API Support

Direct Line provides a REST API that allows external applications to communicate with the bot using REST endpoints. However, there is no API specifically designed to drive the conversation.

Dialogflow, on the other hand, offers a GraphQL API that provides full programmatic control over chat-bot interactions. It also provides client libraries for various programming languages, including Node.js, Swift, Kotlin, and .NET.

Pricing Model

Copilot Studio offers a pricing model based on session usage and integration with Microsoft services. This can be costly for enterprises that rely on multiple Microsoft tools.

Dialogflow, on the other hand, offers a flexible pricing model that includes a free tier and charges based on API calls. This makes it scalable for different use cases.

Additional Features

? include pictures, videos, voice data, streaming, partial responses, conversation history, and custom controls in your conversations.

Copilot offers a range of additional features, including RAG-based capabilities, the ability to use custom sites or SharePoint for RAG, and the option to use alternative generators via hooks.