Microsoft Copilot has been in widespread enterprise deployment for about six months now. The initial hype has settled, and we have enough data to assess what’s actually happening.

I’ve been tracking Copilot adoption across fifteen Australian enterprises. Here’s what I’m seeing.

The Adoption Numbers

Across the organisations I’m tracking:

• Average rollout: 35% of knowledge workers with Copilot licenses
• Active usage (weekly): 60% of license holders
• Heavy usage (daily): 25% of license holders
• Abandoned: 15% tried it and stopped

These numbers are better than I expected based on historical enterprise software adoption. Copilot is sticky in a way that many tools aren’t.

Where It’s Delivering Value

Document drafting. Starting documents from prompts is consistently cited as valuable. First drafts of emails, reports, and proposals that would take 30 minutes now take 5. The output requires editing, but the starting point is useful.

Meeting summarisation. Teams users who enable meeting summaries are generally pleased. The quality is imperfect but “good enough” for most internal meetings. Major time saver for people in back-to-back meetings.

Excel analysis. Power users are getting significant value from Copilot in Excel. Natural language queries against spreadsheets, formula generation, and data analysis tasks are faster.

Information finding. Searching across Microsoft 365 content via Copilot surfaces relevant documents faster than traditional search. Particularly valuable in large organisations with massive content volumes.

Where It’s Underwhelming

Complex writing. For anything requiring nuance, domain expertise, or specific voice, Copilot output requires extensive editing. The time saved is often consumed by revision.

PowerPoint generation. The most disappointing feature. Copilot-generated slides are generic and rarely usable without significant rework. Most users have stopped trying.

Workflow automation. The promise of “just describe what you want” for complex multi-step tasks hasn’t materialised. Power Automate integration is limited.

Specialised work. Domain-specific tasks (legal drafting, technical documentation, financial analysis) need more than Copilot provides. It lacks the specialised knowledge that professionals need.

The Productivity Question

Everyone wants to know: is Copilot making people more productive?

The honest answer: probably, but less than Microsoft claims and harder to measure than you’d expect.

What I’m observing:

Measurable time savings on specific tasks (drafting, summarising) range from 15-30%. That’s meaningful.

Unmeasurable time shifts are harder to assess. Are people doing more or just doing the same amount faster? The answer varies by individual.

Quality impacts are mixed. Some Copilot-assisted work is higher quality; some is lower. The tool enables both.

Adoption effort consumes time initially. The first month of Copilot use probably costs productivity as people learn the tool.

My rough estimate: 10-15% productivity improvement for users who adopt it effectively. Perhaps 25% for power users. Minimal impact for casual users.

The Cost-Benefit Reality

At $45 AUD per user per month, Copilot needs to deliver meaningful value to justify the cost.

For a knowledge worker earning $100,000 annually (roughly $50/hour loaded cost), the Copilot license costs about $540/year. To break even, you need about 11 hours of productivity savings per year – roughly one hour per month.

That threshold is easily met for active users. The question is whether you’re buying licenses for people who won’t use them actively.

My recommendation: pilot with self-selected enthusiasts, measure actual usage, then expand only to employees who demonstrate engagement. Organisation-wide rollouts waste money on inactive users.

Australian-Specific Observations

Data residency is resolved for most use cases. Microsoft’s Australian data centres handle Copilot processing, satisfying most compliance requirements.

Adoption patterns mirror US but with a lag. Australian enterprises are about 3-4 months behind US counterparts in deployment maturity.

Integration with local systems is sometimes challenging. Australia-specific software (accounting packages, government systems) may not connect smoothly.

The Competitive Landscape

Copilot isn’t the only option:

Google Workspace + Gemini is a credible alternative for Google-native organisations. Feature parity is close and improving.

Standalone tools (ChatGPT, Claude) offer more flexibility but less integration. Many enterprises use both – Copilot for Microsoft work, standalone tools for everything else.

Specialised AI tools are emerging for specific functions (legal, coding, analysis). These often outperform Copilot in their domains.

Copilot’s advantage is integration, not capability. If you’re committed to Microsoft 365, Copilot is the obvious choice. If you’re not, alternatives are worth evaluating.

What I’d Recommend

For organisations considering Copilot:

1. Start with a focused pilot (50-100 users) of enthusiasts
2. Measure actual usage, not just satisfaction
3. Identify which roles and tasks benefit most
4. Expand selectively based on evidence
5. Budget for training – untrained users underutilise the tool

For organisations already using Copilot:

1. Track usage data to identify inactive licenses
2. Invest in power user training
3. Share effective prompts and use cases internally
4. Don’t expect PowerPoint generation to improve soon
5. Consider specialised tools for domain-specific needs

The Bottom Line

Copilot is a useful tool that delivers real value for specific tasks and specific users. It’s not the transformative technology that early marketing suggested, but it’s also not the overhyped disappointment that sceptics predicted.

The enterprises getting the most value are those with realistic expectations, selective deployment, and investment in user enablement. Those treating it as magic or ignoring it entirely are both missing the point.

Six months in, the verdict is: worth it, with caveats.