ANZ builds 'enterprise AI control plane' to govern AI agents

ANZ Banking Group reported a cash profit of $7.59 billion for the year ended September 30 — up 31 percent year-on-year — and used its annual results presentation to announce a new centralised layer for governing AI across the bank.

Group CEO Nuno Matos said ANZ has stood up an "enterprise AI control plane" that provides visibility, governance and technical control as the bank deploys increasing numbers of AI agents. The bank disclosed no vendor, product or version details, and iTnews reported receiving no response when it sought comment on whether outside help was involved in the build.

What is the control plane meant to do?

The mechanism is straightforward. The control plane sits above the bank's AI stack and watches everything — users, models, agents, tools, costs and platforms — while recording how decisions were made.

Matos laid out the existing foundations and the new layer in one statement: "We have existing controls in place including multiple AI gateways and a governance council. We are building on this by implementing an enterprise AI control plane which will provide visibility, governance and technical control across users, models, agents, tools, costs and platforms, as well as traceability of decisions."

The control plane adds hard boundaries for autonomous software. "As we increasingly use AI agents, humans must remain in control," Matos said. "Our enterprise control plane will set defined limits on what agents can and do, and stop activity where it operates outside of those limits."

A blog post by ANZ CIO Donald Patra offered no additional architectural detail. What is known is the intent: a kill-switch capability that halts any agent operating outside pre-set limits, plus a cost-and-usage view across every AI system the bank runs.

In plain terms, ANZ has built one place where it can see, and stop, every AI system it operates.

How does this fit the wider pattern?

ANZ is not alone in this approach. Telstra has stood up a similar control plane structure, and it was recently revealed that Accenture played a role in building that capability. ANZ's move makes it the second major Australian enterprise to publicly adopt the pattern.

The timing matters too. ANZ is the third of Australia's big four banks to highlight AI investments in annual results presentations in a little over a week:

• Westpac headlined its AI discussion with a direct dollar saving attribution on a specific project.
• NAB focused on a new staff-facing AI assistant built on technology from Anthropic.
• ANZ led with the control plane, alongside percentage productivity gains from specific AI use cases.

Three banks, three framings — but all three now treat AI governance as results-presentation material.

What is ANZ actually running AI on?

Matos described the bank's adoption approach as "deliberately targeted". ANZ wants AI to deliver a more relevant, personalised experience for customers and better tools for its bankers; to simplify work and lift organisational productivity and capacity; and to strengthen risk management.

"Rather than pursuing AI everywhere, we are prioritising opportunities which unquestionably deliver value and accelerate our strategic initiatives with right-sized investment spend," Matos said.

He set a clear bar for any project: "Our AI use cases require [a] clear business outcome, where the technology can be embedded into the way work gets done and where it can be scaled across the bank. That work is supported by robust governance, which is already in place and controls that we are continuing to strengthen."

Three deployments were named as working examples: an AI-enabled CRM system, AI tooling for the bank's engineers, and AI applications in its contact centres.

The financial backdrop is strong: a 31 percent lift in cash profit — roughly $1.8 billion more than the prior year — gives ANZ room to fund the governance build without pressure.

Governance before scale is the message: ANZ wants its agents on a leash before it lets more of them loose.