50 years of Microsoft: All signs point to Copilot

Shopping, Pages, Podcasts, Search and Deep Research: Microsoft announces a range of new functions for the AI assistant Copilot. After 50 years, the software company seems to have become an AI company. At least that is the main theme of the 50th anniversary celebrations.

Mustafa Suleyman, CEO at Microsoft AI, says he is inspired by the simplicity and power that Bill Gates brought to the PC on every desk all those years ago. "At Microsoft AI, we are driven by the same spirit." Copilot is to become an AI companion for everyone. It is not possible to say exactly what this will look like, as everyone will operate it differently and be operated by it.

Suleyman also announces that the co-pilot will, with the appropriate permission, remember everything you have discussed, what you like and what you don't like. He will keep the name of the dog and the difficult project. Whether this promise goes down so well with everyone is questionable. Not everyone wants to release this data, but there should also be a "permission" – so there is also a way to prevent this. For everyone else, the dynamic and the relationship that people have with a technology is changing. It is a new era.

Copilot: memory, podcasts, research and AI search

The new functions include a memory and a personalization option. The stored data can be edited via a dashboard. Microsoft emphasizes that the user remains in control. Design options for the co-pilot itself have been announced as initial experiments.

Actions refers to the ability that is commonly referred to as AI agent or agentic. The co-pilot can act: buy tickets for events or order a gift for a friend. Copilot Actions should work on most websites. However, there are also specific partners, including Booking.com, Expedia, OpenTable and more. Microsoft is not yet explaining exactly how the AI agent works technically. Similar functions from other providers are generally based on screenshots, which can then be processed by the models in order to recognize input fields, for example, which in turn can be filled in by a model.

Copilot Vision is already available for the web and is now coming to Windows and mobile applications. Behind this is the ability to talk to the Copilot about what can be seen on the screen – either photos or in real time via an open camera. Vision is available via the iOS and Android app, and the new Windows app should also make the function usable in the work environment.

Pages is a kind of canvas, as OpenAI and Google call it. It allows you to load ideas, sketches and more onto a blank canvas. The copilot helps with sorting and editing if required. A podcast function is also already available from competing providers – , such as NotebookLM from Google. Copilot can now create AI-generated podcasts. This means that content can be converted into a podcast by the AI assistant. While listening to it, you can also ask questions about the content.

Shopping is designed to keep an eye on products, provide information about price reductions and enable purchases to be made directly in the app. The agent should be able to do this. However, the functionality is not yet more specific. Deep Research is responsible for complex questions and multi-stage tasks. This means that Copilot can also process larger volumes of content. OpenAI and Google also have a research function up their sleeves.

Copilot Search is the AI search in Bing, although Bing has actually been an AI search for a long time. With the new Copilot function, AI search and traditional search are to be combined and used for a cross-check in order to deliver the best possible results.

(emw)