Digital Economy

Microsoft brings Copilot together in a single super-app: Code and Autopilot are on their way

Microsoft’s aim is to make Copilot the gateway to day-to-day work, software development and agents that carry out tasks over time.

4' min read

Translated by AI
Versione italiana

4' min read

Translated by AI
Versione italiana

Copilot is moving beyond being merely a responsive assistant and is aiming to become a software factory. The unified app is simply called Copilot. In practice, the user doesn’t stop at text or a summary: they move from an idea to a working app, without leaving the environment in which they are working. “Artificial intelligence,” said Microsoft CEO Satya Nadella, “must work for your business, helping it to create and retain more knowledge. It must not reveal what lies at the heart of the company.” ‘Every agent,’ he went on to explain in a lengthy post, ‘must have an identity. Everything it does must be observable. And what we observe must be governable through rules.’

With these announcements, Microsoft is saying that the AI model is no longer the product: it is a component. The value lies in what surrounds it: where the AI-created apps run, who manages them, and who bills for them. Microsoft is playing to its strengths: identity (Entra), business data (Graph and Microsoft 365), IT administration and Azure billing.

Loading...

Why this change?

Let’s start with the example presented by Microsoft. An employee asks Copilot to create an app to coordinate a product launch. The app works. At this point, the less glamorous work begins: deciding where to run it, what data it can access, who can use it and who will keep it up to date. Microsoft’s aim is to make Copilot the gateway to day-to-day work, to software development and to agents that carry out tasks over time.

What do Home, Code and Autopilot do?

Home brings together chats, recent activity and suggestions for picking up where you left off. Microsoft plans to add a control centre for email, calendar, conversations and to-do lists.

Code is also aimed at those who do not code. An employee will be able to describe a requirement and ask Copilot to build a small application or a workflow. Microsoft says the feature will use GitHub technology, an isolated environment for running code and an IT-managed infrastructure for sharing it within the organisation. One practical issue remains: before entrusting these tools with real data and processes, organisations will need to test how they work.

With Autopilot, the user assigns the agent a name, a role and an objective; the agent, hosted in the cloud, can then work in the background without having to wait for a new message. Microsoft presents it as the next step beyond an assistant that only responds when asked. The extent to which it can act independently, and the controls in place over its actions, will be crucial for business use.

Providing context for these features is Microsoft IQ, the system through which the company aims to connect AI to people, documents and ways of working within the organisation. Word, Excel and PowerPoint are integrated more directly into Copilot conversations, whilst keeping files synchronised with traditional applications. In Teams, the assistant will be able to utilise the context and permissions of the chat or channel in which it is called upon.

The real innovation is Copilot Managed Runtime

Apps created by employees run on Microsoft servers, within the company’s Microsoft 365 environment. Identity management, permissions, data access, version control and monitoring are all already included. Previously, an app written by AI was just the start of the work for IT; now, the path to production is built in.

This also applies to code created using third-party tools, provided they use the Microsoft SDK. For IT departments, they explained, it is all about having a single dashboard. All apps, regardless of where they were created (Cowork, Code, Copilot Studio, external tools), appear in a single inventory in the admin centre.

Microsoft’s approach is ‘open build, managed run’: everyone builds wherever they like, but everything runs under the same rules. It is a response to the risk that ‘vibe coding’ might flood companies with apps that are out of control.

Loading...

How much does it cost?

This is the trickiest part, and it’s the one that needs to be understood most clearly. The billing also changes. Microsoft distinguishes between everyday use, which is covered by a monthly licence per user, and the activities of more advanced agents, which are billed on a pay-as-you-go basis.

The new services (Cowork, Code, Autopilot, the runtime and, from October, Copilot Studio agents) are billed on a pay-as-you-go basis, using credits. With Agent 365’s FinOps features, IT departments will be able to manage access and expenditure. The aim is to make AI available to the whole organisation; the real test for buyers will be to work out how much the agents actually cost once they start running non-stop.

How our relationship with Anthropic, OpenAI and others is changing

Microsoft has turned the challenge set out by CEO Satya Nadella – to find its place in the agent economy – into a concrete strategy.

The AI model is no longer the product; it is a component. The value lies in what surrounds it: where the AI-created apps run, who manages them, and who generates revenue from them.

Microsoft is playing to its strengths: identity (Sign In), business data (Graph and Microsoft 365), IT management and Azure billing. These are all things that an AI lab cannot replicate quickly.

For example, Anthropic presents itself as both a partner and a competitor. The option to choose the model family makes Microsoft’s strategy clear: the models may change, whilst the value lies in the platform that manages them. The implicit message to Anthropic and other research laboratories is: ‘You will be suppliers; the customer remains mine’.

Anthropic, for its part, sells directly to businesses. OpenAI is pushing ChatGPT towards agents and the workplace (workspace agents, Codex). Microsoft’s advantage lies not in the model, but in its existing governance framework.

With Google, on the other hand, the competitive challenge is more similar. Google starts with the cloud and Workspace, whilst Microsoft builds on its dominance in productivity solutions for large enterprises. Here, success depends more on winning the trust of CIOs than on the quality of the model.

Apart from the costs, which we are not yet in a position to assess, there is the issue of app control. The code is portable, but governance and execution remain within the Microsoft environment. At a more strategic level, however, the Redmond giant’s advantage holds only as long as the laboratories agree to act as suppliers. If the model regains the upper hand, the balance of power will shift.

Copyright reserved ©
  • Luca Tremolada

    Luca TremoladaGiornalista

    Luogo: Milano via Monte Rosa 91

    Lingue parlate: Inglese, Francese

    Argomenti: Tecnologia, scienza, finanza, startup, dati

    Premi: Premio Gabriele Lanfredini sull’informazione; Premio giornalistico State Street, categoria "Innovation"; DStars 2019, categoria journalism

Loading...

Brand connect

Loading...

Newsletter

Notizie e approfondimenti sugli avvenimenti politici, economici e finanziari.

Iscriviti