ANZ is putting a control plane over its AI agents
Agents change the governance question for a bank. A chatbot answers. An agent acts, and someone has to decide in advance how far it is allowed to go. ANZ's answer, set out with its 2026 full-year results, is an enterprise AI control plane: a layer that gives the bank visibility, governance and technical control across users, models, agents, tools, costs and platforms, with traceability of decisions. Group CEO Nuno Matos said it builds on controls already in place, including multiple AI gateways and a governance council.
Limits before scale
"As we increasingly use AI agents, humans must remain in control," Matos said. The control plane will set defined limits on what agents can do and stop activity that operates outside those limits. Group CIO Donald Patra used the same words in his own account of the year and tied it to a deliberately targeted approach. Rather than pursuing AI everywhere, ANZ says it is prioritising uses that can be embedded into the way work gets done, deliver clear business value and scale across the bank. Patra pointed to three: AI-generated customer summaries inside the CRM for business and private bankers, AI-assisted engineering that he says has doubled capacity to deliver code for the Single Customer Front End program, and transcription, summaries and routing in the Enterprise Contact Centre.
The people half
Patra was explicit that technology alone will not decide the outcome. ANZ is establishing an AI Academy to lift AI literacy across the bank and build role-specific capability "through practical learning, experimentation and accreditation". Rob Amos, a principal software engineer, described AI as a teammate that stops "only when it needs us to apply engineering judgement or make decisions on customer outcomes". Yannie Onsen, head of digital customer engagement, said the aim is for bankers to spend more of their time "in conversations where human judgement and empathy really matter".
What this means for L&D
A control plane can stop an agent at a boundary. It cannot teach a banker where that boundary should sit, or when to override a summary that reads well and is wrong. Those are the decisions the academy's accreditation will have to test, and they belong to the people who own the workflow, not only the platform team. That is where our AI Governance programs start.
Sources:
News and insights for innovation, digital transformation, future of work and L&D leaders.
Stay ahead of learning and development, corporate innovation and digital transformation news. Plus the future of work. For leaders in AU, NZ, HK, SG, the US, the UK and Canada.
