Microsoft 365 Copilot has changed significantly since the early days of:

“Write me an email.”

“Summarise this document.”

“Create a PowerPoint from these notes.”

Those features are still useful.

But Microsoft’s Wave 3 of Microsoft 365 Copilot represents a much bigger shift.

Copilot is moving from being primarily an AI assistant that responds to individual prompts towards something capable of editing your work directly, understanding more of your business context and carrying out longer multi-step tasks through agents.

Microsoft announced Wave 3 in March 2026 as part of a wider push towards what it calls “Frontier Transformation”. The changes span Word, Excel, PowerPoint, Outlook, Copilot Chat and Microsoft’s growing agent ecosystem.

For businesses already using Microsoft 365, this matters because Copilot is becoming less like an optional chatbot sitting alongside your applications and more like an intelligence layer built directly into the way work gets done.

So what has actually changed?

Copilot Can Now Work Directly on Your Existing Content

Earlier AI experiences often felt like this:

Ask the AI something.

Receive a response.

Copy the response.

Paste it somewhere else.

Edit it.

Save it.

Wave 3 aims to remove more of that friction.

Microsoft says Copilot can now work alongside users inside Word, Excel, PowerPoint and Outlook, helping create, edit and refine content directly in the application.

That matters because business work rarely starts with a blank page.

You usually already have:

A Word document.

An existing spreadsheet.

A presentation template.

An email thread.

A customer proposal.

Copilot can increasingly work on those existing materials rather than simply generating another disconnected answer in a chat window.

Word Becomes More Collaborative

In Word, the direction is towards Copilot helping refine an existing document rather than simply producing an initial draft.

That might mean:

Improving structure.

Rewriting particular sections.

Adjusting tone.

Adding missing information.

Using relevant business context.

Reviewing the existing document rather than starting again.

Microsoft has also continued improving citations and grounding in Word so users can better understand where information has come from.

For businesses, that makes Copilot potentially more useful for:

Proposals.

Policies.

Reports.

Customer communications.

Internal documentation.

The productivity gain isn't necessarily writing an entire document for you.

It may be helping reduce the repeated editing that happens afterwards.

Excel Is Becoming Much More Interesting

Excel has traditionally been one of the harder applications for generative AI to work with properly.

Producing a paragraph is relatively easy.

Making safe and useful changes to an existing spreadsheet is considerably harder.

Wave 3 pushes Copilot further towards multi-step editing inside Excel.

Microsoft says Copilot can use Work IQ context to bring in relevant information from emails, meetings, chats and files when working with spreadsheets. It can also edit modern workbooks stored locally on Windows and Mac rather than requiring every supported workbook to live in the cloud.

That opens up some useful possibilities.

For example:

“Add a margin calculation to this table.”

“Highlight products whose sales have fallen over three months.”

“Create a summary sheet using the existing workbook.”

“Explain why this month's figures are different.”

For many businesses, this is where Copilot may move from being interesting to genuinely useful.

PowerPoint Should Understand Your Existing Presentation Better

Anyone who has used AI to generate presentations knows the problem.

You ask for a deck.

You receive something technically correct.

But it doesn't look anything like the presentation your company would actually produce.

Wave 3 is intended to improve that.

Microsoft says Copilot can increasingly understand:

Existing layouts.

Object styles.

Presentation structure.

Brand kits.

and then create or refine slides in a way that better matches the existing deck.

That could make Copilot much more valuable for businesses with established presentation standards.

Instead of generating another generic AI slideshow, you could potentially say:

“Add three slides explaining the new service using the style of this presentation.”

That's a much more practical business workflow.

Outlook Moves Beyond Writing Emails

Email drafting has been one of Copilot's obvious uses from the beginning.

But Wave 3 pushes Outlook further towards understanding the wider context around a conversation.

Rather than asking Copilot to write a generic response, Work IQ can help provide context from:

Meetings.

Previous conversations.

Files.

Relationships.

Other relevant Microsoft 365 information.

The goal is for Copilot to understand more of why you're writing the email, not simply the sentence immediately before it.

That can potentially make email assistance much more useful.

But it also brings us to one of the most important Wave 3 concepts.

What Is Work IQ?

Microsoft describes Work IQ as the intelligence layer behind Microsoft 365 Copilot.

Think of it as the system that helps Copilot understand the context of your work.

Not simply:

“What does this document say?”

But:

Who do you work with?

Which projects are relevant?

What meetings have happened?

Which files are connected?

Which conversations matter?

What is current?

Microsoft says Work IQ has three integrated layers covering data, context, and skills and tools. It draws on information from Microsoft 365 and the relationships between that information to help Copilot produce more relevant results.

That is a major difference from an AI model operating mainly from a prompt and a few uploaded files.

Why Work IQ Matters

Imagine asking Copilot:

“Prepare an update on the new customer project.”

Without business context, an AI system might ask:

Which customer?

Which project?

Where is the information?

With Work IQ, Microsoft wants Copilot to understand that the relevant information may exist across:

An Outlook email thread.

A Teams meeting.

A SharePoint folder.

An Excel workbook.

A PowerPoint.

Several colleagues.

The value isn't simply having more data.

It's understanding the connections between it.

Work IQ APIs Take This Beyond Copilot

Microsoft has also opened this capability to developers through Work IQ APIs, which became generally available in June 2026.

Those APIs allow organisations to build their own agents that can use Microsoft's understanding of work context rather than simply querying raw files or isolated systems.

That becomes particularly interesting for businesses developing their own AI applications or workflow automation.

An agent could potentially understand:

Which project a customer belongs to.

Which employees normally work on it.

Which documents are relevant.

Which meeting happened most recently.

Then use that context when carrying out work.

Microsoft Is Going Multi-Model

Another major Wave 3 change is Microsoft's move towards model diversity.

Microsoft 365 Copilot is no longer positioned as being tied exclusively to one AI model provider.

Microsoft has incorporated models from both OpenAI and Anthropic, including Claude availability in Copilot experiences through its Frontier programme.

From an ordinary employee's perspective, this shouldn't mean:

“Which AI model should I use for this email?”

Microsoft's aim is to make different models available for different types of work while keeping the experience inside Microsoft 365.

That could allow Microsoft to choose models based on things such as:

Reasoning.

Document work.

Analysis.

Long-running tasks.

Performance.

The bigger message is that Copilot is evolving into a platform rather than a wrapper around one particular AI model.

Agents Are Moving Into Everyday Microsoft 365 Work

Another major theme of Wave 3 is agents.

An agent is a more specialised AI capability designed around a particular task, process or information source.

Instead of simply asking general-purpose Copilot questions, you might have an agent dedicated to:

Sales.

HR.

Customer onboarding.

Internal policies.

Project management.

Finance.

Service requests.

Microsoft has been expanding agent creation directly within Microsoft 365, allowing users and organisations to create more focused experiences around their own work.

This doesn't mean every employee should immediately start creating dozens of agents.

It means organisations need to start thinking about AI agents in the same way they think about other business applications:

What is it for?

Who owns it?

Which data can it access?

Who can use it?

How is it secured?

Does it actually improve anything?

Copilot Cowork Is Where Things Get Really Interesting

Perhaps the biggest Wave 3 shift is Copilot Cowork.

This goes beyond:

“Answer this question.”

or:

“Create this document.”

Copilot Cowork is designed for long-running, multi-step work.

You describe the outcome.

Cowork creates a plan.

It can then work through the steps while keeping you involved through checkpoints and approvals.

For example, rather than saying:

“Summarise this spreadsheet.”

you could potentially delegate something more like:

“Review these project files, identify outstanding actions, update the status document and prepare a summary for tomorrow's meeting.”

That's a fundamentally different AI interaction.

You're not simply requesting an answer.

You're delegating a piece of work.

You Still Stay in Control

This is an important part of Microsoft's design.

According to Microsoft, Cowork can:

Create a plan.

Continue working in the background.

Pause when clarification is needed.

Show recommended actions.

Allow users to approve changes before they are applied.

The objective is autonomous execution without simply giving the AI unrestricted control.

Cowork moved into general availability through Microsoft's Frontier programme in March 2026.

Agent 365 Brings Management Into the Picture

If organisations are going to create hundreds or potentially thousands of AI agents, somebody needs to know they exist.

That's where Agent 365 comes in.

Microsoft announced Agent 365 as part of the same Wave 3 strategy, with general availability beginning in May 2026.

The idea is to provide a management layer around agents.

Businesses increasingly need answers to questions such as:

Which agents exist?

Who created them?

Which data can they access?

What actions can they take?

Are they still being used?

Are they secure?

Are they providing value?

That may sound excessive today.

But think about SaaS applications.

Businesses spent years discovering they had dozens of applications employees had independently signed up for.

Microsoft clearly doesn't want agents to become the next version of uncontrolled shadow IT.

The Security Question Gets Bigger With Wave 3

This is perhaps the most important part for SMEs.

An AI assistant that can summarise a document has limited power.

An AI agent capable of reading business information and carrying out actions has much more.

That means AI adoption increasingly becomes an identity, permissions and data-governance problem as well as a productivity opportunity.

Before expanding Copilot use, businesses should understand:

Who can access which SharePoint sites?

Are old Teams sites still accessible?

Do former employees still have permissions?

Is confidential information overshared?

Are sensitivity labels being used appropriately?

Are Microsoft 365 administrator accounts secure?

Which employees can create agents?

What information can those agents access?

AI often doesn't create the underlying permissions problem.

It exposes how messy the permissions already were.

Copilot Can Only Be as Good as the Environment Beneath It

If your Microsoft 365 environment contains:

Years of overshared files.

Poorly controlled permissions.

Old employee accounts.

Duplicate documents.

Unstructured SharePoint sites.

Then giving AI better access to all of that information doesn't automatically improve things.

It can make the consequences of poor information governance more visible.

That's why Hamilton Group recommends fixing Microsoft 365 foundations before rolling Copilot out widely.

The Buff IT Guy Meets Wave 3

The Buff IT Guy has discovered Copilot Cowork.

Naturally, he's immediately interested in delegating leg day.

We've explained that's not how it works.

But there is a useful lesson here.

Early Copilot was a bit like having somebody standing beside you in the gym giving advice.

Wave 3 is increasingly like having somebody capable of taking responsibility for part of the workout.

That is more powerful.

It also requires more control.

Before letting Copilot lift heavier workloads, make sure:

The permissions are right.

The data is organised.

The security is strong.

The users understand what they're doing.

The agents have clear jobs.

Then start adding weight.

Don't simply turn everything on because Microsoft released another impressive demonstration video.

What Should Businesses Do Now?

You don't need to deploy every Wave 3 feature immediately.

We'd recommend a much more controlled approach.

Start With Existing Copilot Use

Are employees already using Microsoft 365 Copilot?

If so:

Which features?

How often?

For what work?

Where is it saving time?

Where is it producing poor results?

Don't introduce another layer of agents before understanding whether employees are getting value from the existing tools.

Review Microsoft 365 Permissions

Before Work IQ and agents become deeply embedded, review access.

Especially:

SharePoint.

Teams.

OneDrive.

Guest users.

Administrator accounts.

Sensitive company information.

The AI should inherit an environment you trust.

Pick a Few Real Processes

Don't begin with:

“How can we use AI?”

Start with:

“What wastes time?”

Maybe:

Preparing monthly reports.

Reviewing project updates.

Finding internal information.

Customer onboarding.

Creating proposals.

Meeting follow-ups.

Then assess whether Copilot, an agent or a custom AI workflow can improve it.

Pilot Agents

Start with a small number of clearly defined agents.

Measure:

How often they're used.

Whether employees trust them.

How much time they save.

Whether output needs correcting.

Whether permissions are appropriate.

Ten useful agents are much better than 500 nobody understands.

Train Your Users

Employees need to understand that Copilot has changed.

A simple prompt such as:

“Write an email.”

doesn't take advantage of what modern Copilot can do.

Training should increasingly cover:

Using work context.

Delegating tasks.

Reviewing AI output.

Understanding agents.

Checking sources.

Protecting confidential data.

Human oversight.

The technology is moving quickly.

User skills need to move with it.

How Hamilton Group Can Help

Hamilton Group can help businesses move from simply having Copilot licences to actually getting useful business outcomes from AI.

That can include:

• Microsoft 365 Copilot readiness assessments
• Microsoft 365 permission reviews
• SharePoint and Teams governance
• Copilot licensing
• Microsoft Entra ID security
• Microsoft Purview and information protection
• Copilot rollout planning
• Employee training
• AI-agent planning and development
• Custom AI-powered business applications
• Workflow automation
• Work IQ and Microsoft 365 integration
• AI governance
• Cyber security
• Ongoing Microsoft 365 management

Hamilton Group is also increasingly able to use AI-assisted application development to create bespoke tools around genuine business productivity problems.

That means the conversation doesn't have to be:

“Which Microsoft AI feature should we switch on?”

It can be:

“What problem are we trying to solve, and what is the best technology for solving it?”

Sometimes the answer will be Copilot.

Sometimes an agent.

Sometimes Power Automate.

Sometimes a bespoke application.

And occasionally the correct answer is:

You don't need AI for this at all.

That's an equally useful conclusion.

Wave 3 Is About Delegation, Not Just Generation

The biggest change in Microsoft 365 Copilot Wave 3 isn't another AI model.

It's the shift from generating things towards doing things.

Copilot can increasingly work directly inside documents.

Work IQ gives it deeper organisational context.

Multiple AI models give Microsoft more flexibility.

Agents allow businesses to create specialised AI workers.

And Copilot Cowork pushes towards delegating longer, multi-step pieces of work.

That has enormous productivity potential.

But it also means businesses need better:

Data governance.

Permissions.

Identity security.

Agent management.

User training.

Oversight.

The stronger the AI becomes, the more important the foundations underneath it become.

The Buff IT Guy would understand that.

Before you load another 40 kg onto the bar, make sure the rack is actually secure.

If your business is considering Microsoft 365 Copilot, Wave 3 or AI agents, Hamilton Group can help you work out where the genuine productivity opportunity is—and make sure the Microsoft 365 environment underneath it is ready.

Call Hamilton Group on 0330 043 0069

Email: hello@hgmssp.com

Visit: hgmssp.com