If you've never used Microsoft Copilot before, you might be mistaken in thinking that it is a product that you can just go out and purchase. Copilot is not a product. It’s a brand name for a set of multiple Microsoft AI technologies. According to Microsoft, the name “Copilot” represents any generative AI experience that a user can have on the Microsoft platform.
To ensure we are all on the same page here, Generative AI refers to artificial intelligence that can create new content, such as text, images, music, or even code, by learning from existing data. Unlike traditional AI, which is designed to recognize patterns and make decisions based on predefined rules, generative AI uses machine learning models to generate new and original content that mimics the patterns it has learned from the training data.
So, it may come as a surprise to learn that Microsoft has more than ten shipping versions of Copilot available, with more to come. Each version is designed to focus on particular tasks and software. To make it easier to understand, we can group the CoPilot versions into six logical buckets.
Version | Description | Price |
Copilot for Windows and Edge | This version integrates with the operating system and browser, providing enhanced control and functionality | $20/Month per user |
Microsoft 365 Copilot | Designed for Office tools, this version helps users with tasks in Word, Excel, and PowerPoint. | $30/Month per user |
Copilot for Business Services | This version supports customer service, ERP, CRM, and Salesforce, making it a valuable business tool. | $50/Month |
Copilot Studio | Aimed at low-code development, this version allows for custom GPT integration | $200/month per organization |
Development Copilot | Specifically for GitHub, this version assists developers with their coding tasks | $21/Month per user |
Copilot Labs and Vision | This version is focused on research and development, offering experimental features and capabilities for future Copilots. |
As we discussed, Copilot is for generative AI experiences. This means that Copilot will use your documents, files, meetings, conversations, and more to learn how you work and build adaptive assistance for generating new or improved documents. Copilot can also take some limited actions. In Copilot for Office, for example, Excel assists with data analysis by formatting data, creating graphs, generating pivot tables, and identifying trends. In PowerPoint, Copilot can create presentations by summarizing information from Word documents and Excel spreadsheets, adjusting text formatting, animation timing, and presentation style based on your prompts.
The next big wave in AI delivery revolves around actions. Actions allow the stringing together of tasks that Copilot can trigger.
In AI, "action" refers to the ability of an AI system to perform tasks or make changes in response to user commands or environmental inputs. This concept is a core component of interactive AI applications, where the AI not only understands information but also takes specific steps based on that understanding.
Here are some ways AI "action" is applied:
Virtual Assistants
(e.g., Microsoft 365 Copilot, ChatGPT with plugins, etc.)
These AI systems can perform digital actions, such as setting reminders, sending emails, generating summaries, or formatting documents based on user prompts.
Automated Process Management
AI can monitor and manage workflows. For example, in customer service, an AI can interpret user requests and trigger actions, like routing a ticket to the right department or automating responses.
Robotics and Autonomous Systems
In physical settings, AI systems in robots, drones, or autonomous vehicles take actions like navigating obstacles, picking up objects, or driving safely on roads.
Generative AI Actions in Content Creation
Tools like Copilot or ChatGPT can produce text, images, code, or even entire documents based on prompts. The action here is generating or modifying digital content.
Machine Learning in Predictive Actions
Machine learning models can analyze data and trigger predictive actions. For example, an AI model could predict equipment failure and automatically initiate maintenance actions.
API-Driven Actions
AI systems can integrate with external APIs, allowing them to trigger actions in other software systems (e.g., updating records in a database, posting on social media, or initiating transactions).
Customizable AI Workflows
Advanced AI platforms allow for user-defined actions. For instance, in Microsoft’s Power Automate, AI can perform tasks across integrated systems based on user-defined workflows and logic.
Learning and Adaptive Actions
Reinforcement learning models learn from their environment to take actions that maximize a cumulative reward. This is particularly useful in scenarios where AI must adapt its behavior over time to perform more effectively.
With these action capabilities, AI goes beyond static responses to enable dynamic, real-time interactions across digital and physical domains.
Eventually, Copilots will be able to work more closely together. For example, an Office Copilot can read an incoming email and invoke a Windows Pilot routine to set up a new virtual environment or install software for a user. The possibilities are endless when these Copilots can interact through actions.
Commenti