For decades, creating software required a specialized skill set.

You needed to understand programming languages, development environments, databases, APIs and deployment processes. Even low-code platforms, while making development easier, still required users to understand how software was structured.

Microsoft now wants to push that boundary considerably further.

The company’s redesigned Copilot includes a new Code capability that allows users to describe an application, dashboard, tracker, automation or workflow in ordinary language. Microsoft says the feature is powered by the same underlying technology used by GitHub Copilot.

That distinction is important.

This isn’t simply another AI chatbot that can produce snippets of programming code.

Microsoft is trying to turn Copilot into a place where a person can describe a business problem and receive a working software solution.

And that could have consequences well beyond Microsoft’s own product ecosystem.

From GitHub developers to Microsoft 365 users

GitHub Copilot was originally aimed primarily at software developers.

Its job was to help programmers write, understand and modify code. GitHub has subsequently expanded Copilot into an agent capable of working on development tasks, including creating changes on branches, running tests and preparing pull requests.

Microsoft’s new Copilot Code takes the underlying technology in a different direction.

Instead of asking a developer to use an AI assistant inside a development workflow, Microsoft is putting the technology inside an application already used by millions of knowledge workers.

A marketing employee could potentially describe a campaign dashboard.

A finance employee could ask for a reporting tool.

A project manager could describe a workflow for tracking tasks.

A small business could create an internal application without maintaining a traditional software-development team for every small requirement.

Microsoft explicitly describes Code as a way for people to build apps, trackers, dashboards, automations and workflows using natural language.

That is a much bigger proposition than AI-assisted programming.

The important change is not “AI can write code”

AI has been generating code for some time.

Developers already use GitHub Copilot and competing systems to generate functions, write tests, explain unfamiliar code and help troubleshoot errors.

The more important development is the changing user interface to software creation.

Instead of:

Idea → requirements → developer → code → testing → deployment

the emerging model becomes:

Idea → natural-language description → AI-generated solution → human review

The developer doesn’t necessarily disappear from that process.

Instead, the developer’s role can move further toward architecture, security, integration, testing and reviewing AI-generated work.

For simpler internal tools, however, Microsoft is betting that some employees may be able to bypass the traditional development process altogether.

Why Microsoft has an unusual advantage here

Microsoft isn’t approaching this market as an AI company alone.

It controls a huge collection of workplace software.

Word, Excel, PowerPoint, Outlook, Teams, SharePoint, Power Platform, GitHub and Azure all sit within Microsoft’s broader ecosystem.

The redesigned Copilot is intended to bring several of those experiences together.

Microsoft says Word, Excel and PowerPoint can now be accessed directly through Copilot, with documents, spreadsheets and presentations remaining synchronized with the corresponding Office applications.

That creates an interesting possibility.

A user might begin with a question in Copilot, turn the result into a spreadsheet, use that information to build a dashboard and then create a small application around the workflow — without moving between a collection of unrelated AI products.

The value isn’t necessarily any individual feature.

It is the connection between them.

Copilot is becoming more than a chatbot

Microsoft’s redesigned Copilot is organized around three major capabilities:

Home

Home brings Chat and Cowork together.

Chat is intended for conversational tasks such as questions, drafting and information retrieval, while Cowork is designed for longer-running work that the user delegates to the system.

Code

Code allows users to describe software solutions using natural language.

Microsoft says users can create applications, trackers, dashboards, automations and workflows, with the resulting software running in a sandboxed environment.

Autopilot

Autopilot, previously known as Scout, goes a step further.

Microsoft describes it as a persistent agent that can continue working on tasks even when the user is away. It can monitor channels, follow up on conversations, run recurring work and resume projects later.

Together, these capabilities suggest that Microsoft is trying to transform Copilot from something people ask questions to into something they delegate work to.

That is a fundamental change in the role of an AI assistant.

The developer’s job could change rather than disappear

The most common question surrounding AI coding tools is whether they will replace software developers.

The more immediate effect may be different.

AI can reduce the amount of time developers spend on routine implementation, but building reliable production software still involves much more than generating code.

Developers need to understand:

• architecture
• security
• authentication
• databases
• scalability
• testing
• infrastructure
• data privacy
• monitoring
• compliance
• integration with existing systems

Those requirements don’t disappear because code can be generated from a prompt.

In fact, they can become more important when software is being created by people who don’t have traditional programming backgrounds.

If more employees can create software, organizations may end up producing more software that needs to be governed.

That could increase demand for developers in areas such as architecture, security, platform engineering and technical oversight even if the amount of manually written code declines.

The biggest opportunity may be small internal applications

There is a category of software that has traditionally been difficult to justify economically.

Imagine a company needs a simple application for:

• tracking equipment
• managing an approval process
• monitoring a marketing campaign
• collecting internal requests
• visualizing a department’s performance
• tracking suppliers
• managing a recurring administrative process

The business may not have enough value in that individual application to justify a conventional development project.

Employees therefore end up using spreadsheets, email chains and manually maintained documents.

AI-generated applications could change that equation.

If someone can describe the requirement and quickly produce a functional internal tool, the cost and time required to create these small applications could fall substantially.

This is where Microsoft’s strategy could have a particularly large impact.

But “anyone can build software” comes with a catch

Lowering the barrier to software development also lowers the barrier to creating bad software.

A technically sophisticated developer may immediately think about authentication, data validation, permissions, backups and failure scenarios.

A business employee creating an application through natural language may not.

That creates an important governance problem.

Microsoft says Code runs in a sandboxed environment and can be hosted within an organization’s Microsoft 365 environment. The company is also introducing the Copilot Managed Runtime, which is intended to provide infrastructure for safely running AI-generated applications under organizational controls.

Those controls will become increasingly important if AI-generated applications start handling real company data.

Security could become the limiting factor

Imagine an employee creates an internal application that works perfectly during a demonstration.

The next questions are much harder:

Who can access it?

What information can it see?

Can it access customer records?

What happens if the employee leaves the company?

Who is responsible for maintaining it?

How is it audited?

What happens when the underlying AI-generated code contains a vulnerability?

These aren’t hypothetical concerns.

GitHub’s own Copilot coding agent has added features such as code referencing so developers can identify when generated code matches public GitHub code and inspect the associated source and licensing information.

As AI-generated software becomes more widespread, organizations will need similar visibility across the applications being created by non-developers.

Microsoft’s answer is governance around the AI

This is why Microsoft’s broader Copilot strategy is important.

The company isn’t only adding more generation capabilities.

It is also building infrastructure around them.

Microsoft says its Copilot Managed Runtime is designed to let applications created through Code and other Copilot experiences run within the organization’s environment, with IT governance. The company is also expanding Microsoft IQ to provide organizational context and introducing a plugin registry for approved capabilities.

This suggests Microsoft sees enterprise AI as more than a collection of chatbots.

The company is trying to build an environment where people, AI agents, applications, data and permissions operate together.

That could become a major part of the competition between Microsoft, Google, OpenAI, Anthropic and other AI platforms.

Usage-based AI changes the economics

There is another important part of Microsoft’s strategy that is easy to overlook.

Microsoft says everyday Copilot capabilities remain covered by the user subscription model, while longer-running agentic capabilities such as Cowork, Code and Autopilot use usage-based billing through Copilot Credits.

This makes economic sense from Microsoft’s perspective.

A conventional chatbot request may require relatively little compute.

An AI agent that spends several minutes or hours using tools, retrieving information, executing code and interacting with applications can consume considerably more resources.

Usage-based billing allows Microsoft to connect revenue more directly to that consumption.

But it also creates a new challenge for businesses.

Companies adopting AI agents will need to understand not just how many employees have access, but how much work those agents actually perform.

Microsoft is already adding FinOps capabilities aimed at helping organizations monitor and control AI spending.

The Microsoft-GitHub connection becomes more important

The strategic importance of GitHub in this transition shouldn’t be underestimated.

Microsoft acquired GitHub in 2018, and GitHub has become one of the company’s most important developer platforms.

Now the technology developed around GitHub Copilot is moving outward.

The path increasingly looks like this:

GitHub → developers → Microsoft Copilot → knowledge workers → AI-generated applications

That expands the potential audience dramatically.

Instead of selling AI coding assistance only to programmers, Microsoft can potentially expose the same underlying technology to everyone who uses Microsoft 365.

That is a very different market.

Does this mean programming becomes obsolete?

No.

It does mean that the definition of programming may continue to change.

For decades, programming meant translating a requirement into instructions that a computer could execute.

Natural-language development introduces another layer.

A person can describe what they want, while an AI system translates that description into software.

But somebody still has to determine whether the result is correct.

That distinction matters.

A generated application can be syntactically valid while being logically wrong.

It can produce the correct output in ordinary circumstances while failing in edge cases.

It can also expose data that the user never intended to make accessible.

So the emergence of AI-generated software may make software literacy more important even as it makes traditional coding less necessary for some tasks.

Microsoft is betting that software creation becomes a normal office skill

Microsoft’s own description of Code makes the ambition unusually clear.

The company argues that the unit of knowledge work has traditionally been the file — the document, spreadsheet or presentation — and that AI-generated “small, purpose-built solutions” could become another fundamental building block.

That is an interesting way to think about the future of workplace software.

Instead of waiting for the IT department to create every specialized tool, individual employees could create applications for their own workflows.

A company could eventually have thousands of small AI-created tools rather than dozens of centrally developed applications.

That could increase productivity.

It could also create a management headache.

The real test will be what happens after the demo

Creating a working prototype is becoming increasingly easy.

The harder question is whether AI-generated applications remain reliable after months of real-world use.

Will employees actually adopt them?

Will they integrate correctly with existing systems?

Will they remain secure?

Will businesses be able to maintain them?

Will AI-generated applications reduce IT workloads or simply create another layer of applications that IT departments have to supervise?

Those questions cannot be answered by the launch demonstration alone.

Microsoft’s rollout is beginning through its Frontier program, with Code and other capabilities expanding over time. Autopilot is moving into private preview at the end of September.

The answers will become clearer as organizations use the tools at scale.

NM Focus Analysis

Microsoft’s move is significant because it changes who gets to participate in software creation.

GitHub Copilot helped developers write software faster.

The new Copilot Code is aimed at helping a much broader group of people create software in the first place.

That distinction could prove more important than the underlying AI model.

If successful, Microsoft’s strategy could turn software development from a specialist function into something closer to a general workplace capability — much like creating a spreadsheet became a normal business skill.

But accessibility brings responsibility.

The easier it becomes to create software, the more important security, permissions, testing, governance and human review become.

The biggest change may therefore not be that AI replaces programmers.

It may be that software becomes something many more people can create, while professional developers increasingly become the people responsible for making sure that software is safe, reliable and scalable.

For Microsoft, putting GitHub’s coding technology inside Copilot is a logical extension of its strategy to make AI part of the entire workplace.

For businesses, however, the more important question will be whether these new tools produce measurable productivity gains without creating an uncontrolled ecosystem of AI-generated applications.

The era of AI-assisted coding is already here. The next phase is AI-assisted software creation for everyone.