Microsoft’s new Copilot experience: what it means and how to prepare

Microsoft’s new work-focused Copilot experience is getting a lot of attention, as you’d expect. Once you strip back the noise, the direction looks fairly obvious.

Microsoft wants Copilot to become the place people start work, rather than another chatbot sitting beside Word, Excel and Teams.

That matters because the conversation is moving on from simply buying licenses and learning how to write better prompts. It’s becoming more about asking for an outcome and letting Copilot find the information, tools and steps needed to get there.

From what Microsoft has shown, the new work-focused experience seems to revolve around three areas.

The New Copilot Experience

Microsoft has launched a new Copilot app experience into a single story:

‘Home’ is the everyday bit for chat, search, creating content and handing work off – this is for delegation or long-running work.

‘Code’ points towards building apps, dashboards and business solutions in plain English.

‘Autopilot’ is the longer-term idea: automated agents that can keep an eye on work, follow things up and act over time, which is a more proactive rather than reactive approach to AI.

These are all grounded across your M365, Dynamics 365 and Power Platform environments in what is known as Work IQ.

Why this shift is important

Home is the practical part now, whereas Code and Autopilot feel more like the direction of travel Microsoft are making.

For me, the important shift is this: instead of choosing an app or agent first, you describe the result you want all within one place and Copilot works out how to help.

We’ve spent decades starting with the application, but that now looks like it is about to change.

Need a document? Open Word. Need figures? Open Excel. Need a meeting? Open Teams. Microsoft is clearly trying to turn that around so you start with the outcome instead.

The new app experience means that Copilot will pull together the apps, information, agents and workflows behind the scenes for you. For example, you might ask it to prepare a customer proposal, pull together the evidence, create a presentation and arrange the review. The point is that a business can stay focused on the result instead of bouncing between separate tools.

The real advantage businesses should look at

Powerful AI models are becoming widely available. Microsoft’s stronger position is that Copilot can work directly within Microsoft 365, such as emails, meetings, documents and calendars, while respecting identity, permissions, organisational knowledge, security, compliance and audit controls. It can also connect to business applications and agents.

💡That feels like the real contest in enterprise AI. The business that gets the most form AI use won’t necessarily have the best model for every single task. It will have the best mix of trusted context, useful actions and proper control.

So what should businesses actually do now?

I don’t think the first question should be, “How many Copilot licenses should we buy?”. Time again, I’ve seen businesses start with this – rather, the better question should be, “How ready is our business for AI-driven work?”

Before increasing the automation, organisations need to sort out the basics: review SharePoint and Teams permissions, deal with old or duplicated information, confirm who owns important data and processes, and put sensible security and retention controls in place. Then choose a few genuinely useful cases and agree how value will be measured. We can all get carried away with the capability of AI, but without some good use cases, AI can feel somewhat redundant, and so some companies reach out for some Copilot consulting help.

Looking at security

When you’ve got a few good use cases, it’s time to look deeper into your security.

Copilot doesn’t create poor permissions, but it can make existing oversharing much easier to find. A messy information estate doesn’t suddenly become well managed because AI sits on top of it.

Before rolling Copilot out more widely, review who can access what and put the right AI guardrails in place to protect confidential information and support compliance.

This becomes even more important as AI moves from simply answering questions to taking actions on your behalf. Agents can be given significant access, AI can still make mistakes, and automated decisions need appropriate oversight. The greater the autonomy you give AI, the more important it is to know what it can access, what it can do and where a human needs to step in.

Looking at cost

Of course, businesses will also need to be thinking about the cost side, which I think has sometimes caught some organisations out.

AI looks to be moving towards a mix of per-user licensing and usage-based charges for longer-running or agent-led work. That means customers will have to manage consumption, similar to how they already manage cloud costs.

Access, budgets, alerts, ownership and reporting can’t be an afterthought. Bluntly, buying Copilot without cost controls and an operating model isn’t an AI strategy. It’s an uncontrolled experiment.

The new cost-consumption model for Microsoft Copilot

Microsoft has announced it is introducing a new way of charging for some of the more advanced AI capabilities, via usage based-billing, between January and March 2027.

Your standard Copilot licence will still cover the everyday Copilot features you use across Microsoft 365. However, some newer AI capabilities we’ve mentioned will work on a pay-as-you-go basis, using Copilot Credits.

This means that, as your organisation starts using more advanced AI features, there could be additional costs on top of your usual Copilot licence.

For new Microsoft Copilot Business licences, Microsoft will enable this consumption-based billing automatically and apply a default monthly spending limit of 4,000 Copilot Credits per user. This is a spending limit rather than a fixed charge, so you only pay for the additional AI capabilities you actually use.

What do these new Copilot Credits mean?

The biggest change is that managing Copilot will no longer just be about how many licences you have. You’ll also need to understand how Copilot is being used and what that usage is costing you.

One of the most common questions is, “What does one Copilot Credit actually get you?”

There isn’t a simple answer to this that will fit every business. The number of credits you use will depend on what you’re asking Copilot to do and the complexity of the task.

That’s why it’s worth setting sensible spending limits to begin with, then experimenting with real tasks and projects. This will give you a much clearer picture of how quickly your organisation uses credits and what that consumption could mean for your overall Copilot costs.

Before rolling out these new Copilot capabilities more widely, it’s worth looking at:

• Who has access to consumption-based AI capabilities
• How much AI your teams are using
• What spending limits you want to put in place
• How you’ll monitor costs as adoption grows
• Whether you’re getting enough value from the additional AI capabilities you’re paying for

The new costing model gives businesses more flexibility to use advanced AI capabilities when they need them, without necessarily paying for them upfront. But it also makes visibility and control over your AI consumption increasingly important.

Looking towards the future

My guess is that more people will begin work in Copilot rather than opening Word, Excel or Teams first.

The familiar apps aren’t going anywhere, but they’ll increasingly sit behind an AI layer that finds information and carries out actions.

I’d also expect managed AI operations to become a normal IT service. Businesses will need ongoing help with agent ownership, permissions, monitoring, security, adoption, cost and retirement, much as they already do with devices, cloud services and cyber security.

The organisations that get the most from this probably won’t be the ones with the greatest number of agents. They’ll be the ones willing to drop weak ideas quickly, protect human judgement and turn the workflows that do work into controlled business standards.

Anyway, the short version is that the new Copilot app isn’t just a redesign. It points towards a different way of working, where people set the outcome and AI coordinates more of what happens next.

This new opportunity is real, but the preparation matters. Clean data, sensible permissions, clear ownership, cost controls and proper governance will do more than another round of generic prompt training.

My advice would be to prepare your business – get the foundations right, prove value in a few controlled cases and only expand AI where the results justify it.