Microsoft 365 Copilot Wave 2, announced at the September 2024 event and rolling out through to mid-2025, represents the largest single expansion of in-product AI capability Microsoft has shipped. The headline features are well-documented. What is less well-documented is which of those features actually pay off quickly for Australian businesses, and which are better treated as 2026 projects once the org has its adoption fundamentals in place.
Frontrow's view, after walking a number of AU tenants through Wave 2 configuration in the months since launch, is that the feature set splits cleanly into two groups: high-immediate-value features that repay the setup cost in weeks, and interesting-but-premature features that require data hygiene, agent governance or org maturity that most tenants won't have in place yet. What follows is the honest split.
The features worth turning on first
Copilot Pages in BizChat
Copilot Pages is the multiplayer output surface that Wave 2 introduced into Microsoft 365 Copilot's chat experience. When a user runs a grounded research query, summarise the last six months of board pack commentary, pull the key themes from the client feedback, the result lands in a Page that colleagues can open, edit and extend in real time. It eliminates the copy-paste step that made Copilot outputs disposable. The setup cost is effectively zero: it is available by default inside Copilot Chat for M365 Copilot licensed users. First 30 days: run one team session showing your exec pilot users how to save outputs as Pages instead of copying them to Word. That single habit change materially improves the stickiness of Copilot outputs.
Copilot in Teams meetings, enhanced transcript and task extraction
Wave 2 sharpened the in-meeting Copilot capability significantly. Beyond transcription, Copilot can now surface action items in real time mid-meeting, not just at the end, and the post-meeting summary now includes a structured task list that cross-references who said what. For Australian organisations where meeting volume is high and follow-through is inconsistent, this is the single highest-ROI Wave 2 feature because it replaces a manual task that happens after every meeting. Configuration: ensure meeting recording and transcription policies are enabled in the Teams admin centre and that your Copilot licences have the meeting features policy applied.
Researcher and Analyst agents in BizChat
Wave 2 introduced two named agents into Copilot Chat: Researcher, which can perform deep research tasks using both your organisational data and Bing, and Analyst, which can write and execute Python code against uploaded data files without leaving the chat interface. Both are available inside Copilot Chat under the 'Agents' menu for M365 Copilot licensed users. First 30 days: assign three to five use cases each to Researcher and Analyst across your pilot group. Researcher works well for competitive intelligence and policy summarisation. Analyst works well for quick data cleaning and pivot generation without opening Excel. Measure the hours displaced, not user satisfaction scores.
What can wait
Copilot Studio agent publishing, Copilot extensibility connectors, and Narrative Builder in PowerPoint are all credible features. Frontrow does not recommend activating any of them in the first 30 days of a Wave 2 deployment unless an org already has a mature Copilot Studio governance model in place.
Agent publishing in particular requires a clear policy on who can create and publish agents, what data they can access, and how agent outputs are reviewed. Without that governance baseline, enabling Copilot Studio broadly inside a tenant creates a shadow-IT risk that is harder to walk back than it is to prevent. The September 2025 Copilot Studio updates gave admins the controls to manage this properly; they also assume the admin team knows they need to configure them.
The AU-specific consideration: data residency before features
Several Wave 2 features, Researcher in particular, given its Bing grounding, route processing through Microsoft's global AI infrastructure before data residency commitments kick in. For regulated Australian organisations, verifying the data processing boundaries for each Wave 2 feature before enabling it is not optional. The Microsoft admin centre's privacy and data management settings now carry per-feature data handling notes. Review those before an org-wide enablement.
Try it
Model the Wave 2 return
Adjust the hours-saved inputs per role to reflect Copilot Pages, meeting task extraction and Analyst agent use cases. Export the updated business case for your next review.
Assumptions
Tune your Copilot business case.
Roles
Live result
$704,668
Net annual benefit
- Active users
- 73
- ROI
- 1788%
- Hours / year
- 8,786
- Payback
- 0.6 mo
- Value saved
- $744,088
- Licence cost
- $39,420
Directional only. Real outcomes depend on licence mix, adoption and which workflows you actually target. Book a review to ground the model against tenant telemetry.
Role-by-role breakdown
| Role | Active | Hours/yr | Value | Licence | Net |
|---|---|---|---|---|---|
| Leadership / Exec | 5 | 920 | $143,000 | $2,700 | $140,300 |
| Managers | 14 | 1,932 | $191,100 | $7,560 | $183,540 |
| Knowledge workers | 42 | 4,830 | $324,187 | $22,680 | $301,507 |
| Sales & client-facing | 12 | 1,104 | $85,800 | $6,480 | $79,320 |