Microsoft AI Tour – Key Takeways

The headline announcement was an expansion of Microsoft's digital skilling commitment: a target to train an additional 200,000 New Zealanders in AI skills by 2028, doubling the pledge made in December 2024. Microsoft-commissioned EY-Parthenon research accompanied this, projecting generative AI could contribute between $76 billion and $108 billion to the New Zealand economy by 2038.

A More Coherent Architecture

The more substantive theme across the sessions I attended was a shift in how Microsoft is structuring its AI offering. Over the past couple of years, AI capabilities were added product by product - a Copilot in Teams, another in Power BI, another in Outlook - each reasonably capable in isolation but without a unifying layer beneath them.

The current direction addresses this. Microsoft has introduced three named intelligence layers that sit across the stack:

• Work IQ - the reasoning and memory layer for M365 Copilot, covering user data, workflow habits, and next-action inference
• Fabric IQ - the equivalent for analytical and operational data within Microsoft Fabric and Azure Databricks
• Foundry IQ - a knowledge grounding system within Azure AI Foundry, intended to anchor agents to enterprise data sources

These aren't new products so much as a structural reframe of existing capabilities. The practical effect is that M365 productivity, the data platform, and the developer toolchain are now described under a single architectural narrative rather than as parallel product lines. Whether the integration holds up in practice will take time to assess, but the intent is clear.

Agents as the Operative Model

The sessions consistently used the "Frontier Firm" framework to describe how organisations are expected to move through AI adoption: from Copilot as an individual productivity assistant, to human-led agent deployment for specific workflows, to largely agent-operated processes with human oversight. The framing is Microsoft's own, but the underlying trajectory - AI handling more sequential, multi-step work rather than one-off queries - reflects a broader industry direction.

Concrete product milestones here include the general availability of the Azure AI Foundry Agent Service, which allows multi-agent orchestration in enterprise environments, and M365 Copilot Cowork, currently in early access, which extends Copilot to handle longer-running tasks across files and conversations rather than single-turn interactions.

Copilot Studio featured prominently as the low-code bridge for organisations that want to build agents without deep Azure development capability; the demos ranged from simple retrieval agents to multi-agent patterns connecting to backend systems via Logic Apps.

The overall impression from the day is less about any single announcement and more about consolidation: Microsoft bringing a coherent architecture story to a product portfolio that had grown somewhat diffuse. The agentic direction is not unique to Microsoft. It is where most enterprise AI platforms are headed. But the depth of integration across productivity, data, and developer tooling gives the Microsoft stack some structural advantages for organisations already in the ecosystem.