Six months ago, many Australian enterprises deployed Microsoft 365 Copilot broadly. The early hype promised 30-40% productivity gains. What’s actually happened?

I’ve been collecting data from clients and industry contacts who’ve deployed Copilot at scale. Here’s what I’m seeing.

Adoption Patterns

Initial enthusiasm, then settling. Week one saw high usage as employees experimented. By month two, usage patterns stabilised. Many employees stopped using it entirely.

Power users emerged. Approximately 15-20% of users became regular, skilled Copilot users. These employees genuinely find it valuable and use it daily.

Feature concentration. Most value comes from a few features: meeting summarisation, email drafting assistance, and document summarisation. More advanced capabilities (Excel analysis, PowerPoint generation) see limited adoption.

Training matters enormously. Organisations that invested in training see higher adoption. Those that just deployed licenses see lower engagement.

Where It’s Actually Valuable

Based on real usage, Copilot delivers genuine value in specific scenarios:

Meeting Management

Summarising meetings, generating action items, and catching up on missed meetings. This is consistently rated as the most valuable capability.

Why it works: The problem is real (too many meetings, poor note-taking), the input is structured (Teams recordings), and the output is useful (summaries you can skim).

Caveat: Requires meetings recorded in Teams. Doesn’t work for external calls or offline meetings.

Email Triage

Summarising long email threads, drafting routine responses, and prioritising inbox review.

Why it works: Email overload is universal. Summarising a 50-message thread saves real time.

Caveat: Quality depends on email clarity. Confusing threads produce confusing summaries.

Document Summarisation

Summarising long documents to extract key points before detailed reading.

Why it works: Most business documents are longer than necessary. Getting the gist first helps prioritise attention.

Caveat: Summarisation loses nuance. For important documents, you still need to read the detail.

First Draft Generation

Starting documents, emails, or presentations with AI assistance rather than blank pages.

Why it works: Starting is often harder than editing. A mediocre first draft you improve is faster than starting from nothing.

Caveat: Outputs require significant editing. “Generate a presentation” produces something you’d be embarrassed to present.

Where It Disappoints

Several promised capabilities haven’t delivered as expected:

Excel Analysis

The promise of asking natural language questions about spreadsheets hasn’t materialised for complex analysis. It works for simple queries (“what’s the total?”) but struggles with nuance.

The gap: Real Excel work involves understanding data context, catching anomalies, and knowing what questions to ask. Copilot can answer questions but can’t think analytically.

PowerPoint Generation

“Create a presentation about X” produces bland, generic slides. Useful as rough drafts for internal meetings, not suitable for client-facing work.

The gap: Good presentations require understanding audience, crafting narrative, and making editorial choices. Copilot generates content, not communication.

Search Across Content

Finding information across Microsoft 365 content remains difficult. Copilot helps but doesn’t solve the fundamental problem of poor information architecture.

The gap: If your SharePoint is a mess, Copilot helps you search the mess. It doesn’t organise the mess.

Complex Workflows

Multi-step work that spans applications doesn’t flow as smoothly as demos suggested. Each step requires human oversight and correction.

The gap: Demonstrations show best-case scenarios. Real work involves exceptions, edge cases, and context Copilot doesn’t have.

Measuring ROI

The ROI question is harder than it appears:

Time savings are real but diffuse. Users report saving 15-30 minutes per day. But this time disappears into other work rather than being recaptured visibly.

Productivity metrics don’t move. Most organisations haven’t seen measurable productivity metric improvement (emails processed, documents produced, etc.). The time savings don’t aggregate into visible output changes.

Quality improvements are unmeasured. If Copilot helps produce better meeting summaries, how do you measure that? Most organisations aren’t trying.

Cost is concrete, benefits are fuzzy. $30/user/month is clearly measurable. The benefit is… vague feelings of efficiency.

At scale, the cost is significant. 1,000 users is $360,000 annually. Is that delivering $360,000 in value? Honest answer for most organisations: unclear.

Making Copilot Work Better

Organisations getting more value share certain practices:

Invest in training. Not just “how to use Copilot” but “how to prompt effectively” and “which scenarios benefit most.” Skilled prompting dramatically improves outputs.

Identify and support power users. Find the 15-20% who use it effectively and have them share practices. Peer learning is more effective than formal training.

Set realistic expectations. Copilot is an assistant, not a replacement. Position it as “helping with routine work” not “transforming productivity.”

Optimise underlying data. Copilot’s quality depends on your content quality. If your SharePoint is disorganised, clean it up.

Measure what you can. Even if comprehensive ROI is hard, measure adoption, user satisfaction, and use case frequency. Data helps improvement.

The Licensing Question

At six months, many organisations are reconsidering licensing scope:

Full deployment is expensive for actual usage. If only 20% use it regularly, paying for 100% is wasteful.

Targeted deployment to high-value roles may make more sense. Knowledge workers who write extensively, attend many meetings, or manage information overload benefit most.

Waiting for price reductions is a reasonable strategy. Competition will likely drive prices down. Deferring broad deployment until economics improve isn’t unreasonable.

Vendor vs. Reality

Microsoft’s messaging emphasises transformational productivity gains. Reality is more modest:

Microsoft Claim	Reality
30-40% time savings	10-15% for active users
Universal applicability	Strong for specific use cases
Transforms work	Incrementally improves routine tasks
Easy deployment	Requires training and change management
Clear ROI	ROI difficult to measure concretely

This isn’t to say Copilot has no value – it does. But expectations set by vendor marketing exceed what most organisations experience.

Final Thought

Six months in, Microsoft Copilot is a useful tool for specific tasks, not a productivity transformation. Organisations getting value are those with realistic expectations, investment in training, and focus on high-value use cases.

The 30-40% productivity revolution hasn’t arrived. The modest-but-real productivity improvement has. Whether that justifies the cost depends on your organisation’s specific context.

Evaluate based on your actual usage and value, not vendor promises. That’s the honest assessment that should drive licensing decisions going forward.