IBM and Microsoft's agent platform: what NZ Microsoft shops can use

What IBM and Microsoft announced

On 10 November 2026, IBM and Microsoft announced Imagine Enterprise, a joint platform for building, running and governing AI agents that carry out industry-specific business processes. IBM's part is its Consulting Advantage delivery platform, with tools called Process Studio, Context Studio and Advantage Core, plus more than 6,000 prebuilt agents and assets. Microsoft supplies Foundry, Copilot Studio and GitHub Copilot to build and coordinate the agents. It also supplies the foundation underneath: Fabric for data, Entra for identity, Purview for governance, Defender for security, and Azure for hosting.

The first versions target financial services, insurance and retail. According to the release, manufacturing, energy, the public sector and mobility will follow later on a "repeatable blueprint". Most customers will have IBM Consulting deploy and operate the platform for them or build with them, although self-managed options also exist. Both companies are adding forward-deployed engineers for qualifying engagements. The release gives no pricing, general availability date or list of regions. The only customer it names is SEI, a US financial services firm, and the release describes SEI as exploring agentic technology with the two companies. SEI hasn't announced a production deployment.

CIO Dive's coverage puts the launch in context. Google, AWS, Microsoft and IBM have all started forward-deployed engineering programmes because scaling agents past the pilot stage is proving hard. CIO Dive also cites Gartner research finding that only 4% of CIOs have put plans in place to adjust roles for a workforce of people and agents working together.

Why the launch matters beyond big banks

Few NZ organisations will buy Imagine Enterprise. The platform is built around large consulting engagements, and the first industries are ones where IBM already has big global clients. The architecture is still worth studying, because most of it runs on products NZ businesses already license. If your firm uses Microsoft 365 Business Premium or an E3 or E5 plan, you already have some level of Entra, Purview and Defender, and many of you already have Copilot Studio.

IBM and Microsoft have made the same bet that runs through most enterprise agent announcements this year. In their view, the agents are the easy part, and the hard part is the layer around them: who an agent acts as, which data it can read, what it's allowed to change, and how anyone finds out when it goes wrong. IBM chief executive Arvind Krishna said as much on IBM's October earnings call, as CIO Dive reports: "AI is only as effective as the system around it." The platform puts identity, governance and security tooling at the bottom of the stack and builds the agents on top.

We think NZ firms should read the 6,000 prebuilt agents with care. Prebuilt agents carry assumptions about how a process runs, and IBM's assumptions come from large overseas financial services and retail clients. A claims agent designed for a US insurer won't know about ACC, and a retail agent won't know your Foodstuffs or Woolworths supplier terms. Prebuilt agents give you a starting point, but most will need real configuration before they match how a NZ business works, and that configuration is where the consulting hours go. The release says the platform plugs into existing data, systems and processes. Even so, nothing published so far shows how much of that integration work customers will pay for themselves.

What NZ businesses can take from it

The useful part for mid-sized NZ organisations is the order of work the platform implies: identity and data controls go in before agents are allowed to act. In our experience, many NZ businesses run Copilot Studio pilots before they've settled who owns an agent, what it can reach, or how it's switched off. You don't need IBM or Microsoft engineers on site to fix that order. Most of the controls sit in licences you already pay for, and getting them set up properly is a few weeks of focused work.

Each agent should have its own identity in Entra, with an owner, a stated purpose and an expiry review. Sharing a service account between several automations makes their actions impossible to audit later. Purview sensitivity labels decide which content an agent can surface, so a SharePoint tenant full of unlabelled HR and finance folders is a risk whether you deploy one agent or fifty. Defender and the Microsoft 365 audit log are where you'd find out an agent did something unexpected, and only if someone actually reviews them.

Vendor dependence is the other point to weigh. A platform built jointly by two vendors and run by one consultancy ties your agent logic, process models and governance to that combination. If your firm goes down a similar route with any large partner, find out before you sign whether you can export your agent definitions, prompts and process maps in a form you control. The Imagine Enterprise release doesn't address portability.

Smaller organisations should also notice the industry ordering. Financial services, insurance and retail come first because of their transaction volumes and regulatory pressure. NZ firms in those sectors will probably see their larger competitors, or the Australian parents of NZ subsidiaries, test platforms like this over the next year. Firms in manufacturing, energy or the public sector can expect a longer wait, which gives them time to sort out their data before agents arrive in their sector.

Practical next steps

• List every agent and automation running in your Microsoft tenant, including Copilot Studio agents and Power Automate flows created by staff. Give each one a named owner.
• Move agents off shared service accounts onto their own Entra identities with least-privilege permissions, and schedule a review of each identity every quarter.
• Label your sensitive SharePoint and OneDrive content in Purview before you widen any agent's access. Start with HR, finance and customer records.
• Pick one well-defined process with clear inputs, such as invoice coding or ticket triage, and pilot one agent there with human approval on every action that changes data. Our AI agents service page covers how we scope pilots like this.
• Set up a weekly review of agent activity in the audit log, and document how to disable each agent quickly.
• Before you sign with any platform vendor, ask in writing about pricing per agent or per action, data residency for NZ, and how to export your agent configurations if you leave.
• Write a short agent policy covering who can create agents, what they can touch, and when a person must approve an action. Our AI consultancy team can help if you'd like a second view on governance.

IBM and Microsoft haven't said when Imagine Enterprise will reach industries beyond the first three or whether it'll be offered in Australia and New Zealand. We'll watch for regional availability and the first production case studies. The groundwork above is worth doing either way, because it's the same work any serious agent deployment depends on.