AI Is Learning to Work: Microsoft Copilot Moves from Assistant to Agent
Microsoft Copilot Is Learning to Work
Are We Ready for AI That Acts?
For the past few years, our relationship with artificial intelligence has followed a fairly simple pattern.
“Write this email.”
“Summarise this meeting.”
“Analyse this spreadsheet.”
“Create a presentation.”
And then we take over again.
But what happens when AI does not wait for the next prompt? What happens when we can give an AI system an objective, allow it to work across our business tools, and have it continue working after we have moved on to something else?
On 25 September 2026, Microsoft unveiled a major evolution of Copilot built around Home, Code and Autopilot. The direction is significant: Copilot is increasingly being designed not only to respond to requests, but also to build solutions and carry out longer-running work.
From Prompt to Action
Think about how most people use AI today. You give ChatGPT, Copilot or another AI tool a prompt. It processes your request and produces a response. You remain responsible for moving the work forward.
An AI agent changes that relationship.
Instead of:
“Draft an email asking everyone for their feedback.”
We move toward:
“Coordinate the supplier review, follow up with the relevant people and keep the process moving.”
The first request produces content. The second delegates an objective. That distinction is at the heart of what Microsoft is building.
Home: Bringing AI and Your Work Together
Microsoft’s redesigned Home brings Chat and Cowork together while integrating familiar Microsoft 365 applications more directly into Copilot.
Word, Excel and PowerPoint are becoming part of the Copilot experience itself. Instead of asking AI for information and then copying the result into another application, Microsoft’s vision is that Copilot can help create and update the actual documents, spreadsheets and presentations where work happens.
Code: What If You Could Simply Describe the Tool You Need?
Imagine running a small business and needing a system to track enquiries. You do not know how to code. Traditionally, you might buy software, hire a developer or try to build something yourself using several different tools.
Microsoft’s new Code experience is designed to let users describe what they need in natural language. That could be an app, dashboard, tracker, automation or internal workflow.
This does not eliminate the need for developers. Complex systems still require expertise in architecture, security, testing, integration and maintenance. But it may shorten the distance between an idea and a working solution.
Then Comes Autopilot
Microsoft describes Autopilot as a persistent, proactive and personal agent. It is designed to perform longer-running work rather than simply complete one request and stop.
Microsoft says Autopilot can monitor channels, follow up on conversations, perform recurring work and return to projects over time. Because the work is cloud-hosted, tasks can continue while the user is doing something else.
Your New Colleague Might Be an AI Agent
Imagine arriving at work in the morning. While you were away, an authorised AI agent has reviewed project information, prepared documents, analysed data, monitored designated conversations, followed up on certain tasks and prepared material for your next meeting.
You do not necessarily start your morning by telling the AI what to do. You start by reviewing what it has already done.
That begins to resemble delegation. And delegation requires something that prompt writing does not: management.
But Who Gives the AI Permission?
An AI that writes a paragraph presents one level of risk. An AI that can take actions inside a business environment presents another.
Microsoft says Autopilot is designed to operate inside an organisation’s tenant with its own identity, memory, computer and workspace, alongside organisational permissions, auditing and governance controls.
What is this AI allowed to see?
Which systems may it access?
Which actions can it perform without approval?
Who checks its work?
What happens when it makes a mistake?
Who remains accountable for the final outcome?
AI That Becomes More Powerful Needs Stronger Guardrails
There is an understandable temptation to think that more autonomous AI means humans will need to do less. In some areas, that may be true. But greater autonomy also creates greater responsibility.
If we delegate work to people, we do not simply give them unlimited access to everything in the organisation. We establish roles, permissions, responsibilities, approval processes and oversight. AI agents will require similar thinking.
A better question is “What should we automate, under what conditions, and with what human oversight?”
The Most Valuable AI Skill May Not Be Prompt Engineering
For several years, we have been told that everyone needs to learn how to write better prompts. Prompting certainly matters. But the age of AI agents may require something considerably broader.
If AI can increasingly execute work, employees will need to learn how to define objectives clearly, delegate appropriately, design workflows, establish boundaries, evaluate results, identify errors, protect sensitive information and know when human judgement must override automation.
There Is Another Question Businesses Cannot Ignore: What Does Autonomous AI Cost?
Microsoft says everyday Copilot interactions such as answering questions, summarising information and drafting content can remain associated with user licensing, while longer-running agentic workloads involving Cowork, Code and Autopilot use usage-based billing.
This means businesses may eventually need to evaluate AI workloads just as they evaluate other operational investments: What does this process cost? How much employee time does it save? Does it improve the outcome? Is automation actually cheaper?
A Reality Check: This Future Has Not Fully Arrived Yet
The announcement does not mean that every Microsoft 365 customer can suddenly deploy fully autonomous Copilot agents today. Microsoft says Home and Code are beginning their rollout through its Frontier programme, while Autopilot is moving into private preview. Availability will expand over time.
Businesses therefore should not confuse Microsoft’s direction of travel with universal availability. But direction matters.
From Using AI to Managing AI
Perhaps we have been asking the wrong question.
For years, organisations have asked: “How do we get our employees to use AI?”
Soon, the more important question may be: “How do we prepare our employees to supervise AI?”
Because when AI begins performing work rather than merely discussing it, the employee’s role changes too. We become reviewers, decision-makers, workflow designers and supervisors.
The Skunkworks Academy Perspective
At Skunkworks Academy, we believe AI literacy has to grow alongside AI capability.
Learning AI cannot simply mean learning which button to press or memorising the perfect prompt. It means understanding what the technology can do, what it should do, what it should not do, and where human judgement remains essential.
Microsoft’s latest Copilot announcement gives us a glimpse of the next stage. AI first helped us find information. Then it helped us create. Now it is beginning to act.
Explore Microsoft 365 with Copilot for Business
Interested in Microsoft 365 with Copilot for your organisation? Explore the options below and contact our sales team for product enquiries.
sales@skunkworks.africa
Maria José Adão-Dercksen
Maria José Adão-Dercksen is part of the Skunkworks team, where she focuses on training, digital skills and helping people understand how emerging technologies can be applied in the real world.
With a strong interest in artificial intelligence, AI literacy, Microsoft technologies and the future of work, Maria José writes about technological change from a practical, human perspective—looking beyond the latest tools to explore what they mean for businesses, employees and everyday life.
Through Skunkworks Academy, she advocates for continuous learning and believes that preparing for an AI-powered future is not simply about adopting new technology. It is about developing the knowledge, confidence and critical-thinking skills to use that technology responsibly.
Website: www.skunkworksacademy.com
Blog: blog.skunkworksacademy.com
Training & Enquiries: training@skunkworksacademy.com
Sources & Further Reading
Microsoft — 25 September 2026: Introducing the new Copilot with Home, Code and Autopilot
The Next Web — 25 September 2026: Microsoft’s new Copilot app adds coding and an always-on Autopilot agent
Comments
Post a Comment