In global banking, the AI race is being won by those who are best governedAI is already moving rapidly through the banking system. Agentic AI, capable of taking actions with greater autonomy, promises to push adoption considerably further.Elea Wurth, Max Murray and Michael WilliamsTechnology is moving faster in banking than the institutions governing itGift this article4 min read2 hours agoBanks have spent the past few years proving that AI can work. The harder task has begun – making it work at scale.AI is already moving rapidly through the banking system. The proportion of bank employees using AI at least weekly has more than doubled in the past year, from 30 per cent to 63 per cent. Most large global banks have deployed AI across functions ranging from customer service and technology to marketing, operations and finance. Agentic AI, capable of taking actions with greater autonomy, promises to push adoption considerably further.That distinction matters because the technology is moving faster than the institutions governing it.Dr Elea Wurth, Lead Partner, Trustworthy AI, Deloitte Asia Pacific and AustraliaDeloitte’s Banking on Trust study, based on interviews and surveys with 135 senior banking leaders from Global Systemically Important Banks, Domestic Systemically Important Banks and other large banks, finds a striking gap between AI ambition and governance maturity.The Trustworthy AI Governance Index assessed banks across four maturity levels: Ad hoc (with little or no AI governance); Rudimentary, (emerging but inconsistent); Established, (defined and partially integrated); and Optimised, (supporting strategic objectives and embedded across culture and operations). Some 87 per cent of banks still have substantial room to improve. Only 13 per cent have reached an optimised level, while 10 per cent are operating effectively without AI governance.As Dr Elea Wurth, lead partner, Trustworthy AI, Deloitte, observes, “The question is no longer whether banks can innovate with AI. It is whether they can govern AI well enough to scale it safely and commercially.”For years, governance has often been treated as the natural enemy of innovation: the function that arrives late, raises awkward questions and adds another assessment before deployment. Yet the evidence suggests the opposite may be true. Poor governance may allow experimentation to proliferate. Good governance is what allows it to scale.Max Murray, Head of Financial Services, Deloitte AustraliaDeloitte’s analysis finds that banks with stronger governance scores also tend to report stronger revenue growth, greater deployment and use of AI across business functions. Its modelling suggests that a ten-point increase in governance maturity is associated with a 10 per cent revenue growth.Correlation is not causation. Better-run banks may simply be better at both governance and growth. But the relationship points to an important commercial reality. The value of AI depends less on the brilliance of an individual model than on an institution’s ability to deploy AI consistently across the enterprise, integrate it into business processes and manage the resulting risks.Deloitte Australia’s head of financial services, Max Murray, echoes the lesson that, with AI, technology alone is not enough: “Financial institutions do not need more isolated AI pilots. They need operating models that make responsible AI repeatable across the enterprise.”The costs of getting this wrong are becoming harder to ignore. Financial services report more AI-related incidents and hazards than any other industry. In the first half of 2026, the sector recorded more incidents than in all of 2025, while annual incidents had already risen nearly eightfold between 2022 and 2025.The obvious costs are financial loss, remediation, regulatory scrutiny and reputational damage. A quieter cost of poor governance is paralysis. Where accountability is unclear or controls are cumbersome, executives can become reluctant to approve AI at scale. Institutions then oscillate between excessive caution and hurried deployment once competitive pressure becomes impossible to ignore.For banks, the consequences are amplified because trust is not merely a reputational asset; it is part of the product. AI may make fraud detection faster, service cheaper and financial decisions more sophisticated. But one badly governed system can quickly erode customer confidence. Michael Williams, Head of Banking, Deloitte AustraliaAs Michael Williams, head of banking, Deloitte Australia, observes, “Banking has always been based on trust. The banks that outperform over the next decade will be the ones that innovate and grow while earning and maintaining trust with customers, regulators and markets.”Agentic AI makes this challenge harder. Traditional governance assumes humans review key decisions, but that assumption weakens as AI systems can plan, decide, and act autonomously. Banks will need to move from “human in the loop” towards “human on the loop”, with people setting boundaries, monitoring behaviour, reviewing exceptions and intervening when needed.Yet only 44 per cent of banks report risk monitoring across the implementation lifecycle for agentic AI, compared with 61 per cent for traditional AI and 59 per cent for generative AI. Many may therefore be strengthening governance for yesterday’s AI while tomorrow’s systems are already arriving.The winners will not be the banks that move fastest first. It will be those that build enough trust to keep moving.Dr Elea Wurth is lead partner, Trustworthy AI, Deloitte Asia Pacific and AustraliaMax Murray is head of financial services, Deloitte Australia Michael Williams is head of banking, Deloitte Australia Disclaimer     This article contains general information only, and none of Deloitte Touche Tohmatsu Limited (“DTTL”), its global network of member firms or their related entities (collectively, the “Deloitte organisation”) is, by means of this communication, rendering professional advice or services. Before making any decision or taking any action that may affect your finances or your business, you should consult a qualified professional adviser. No representations, warranties or undertakings (express or implied) are given as to the accuracy or completeness of the information in this communication, and none of DTTL, its member firms, related entities, employees or agents shall be liable or responsible for any loss or damage whatsoever arising directly or indirectly in connection with any person relying on this communication.About Deloitte  Deloitte refers to one or more of Deloitte Touche Tohmatsu Limited (“DTTL”), its global network of member firms, and their related entities (collectively, the “Deloitte organisation”). DTTL (also referred to as “Deloitte Global”) and each of its member firms and related entities are legally separate and independent entities, which cannot obligate or bind each other in respect of third parties. DTTL and each DTTL member firm and related entity is liable only for its own acts and omissions, and not those of each other. DTTL does not provide services to clients. Please see www.deloitte.com/about to learn more.The Australian partnership of Deloitte Touche Tohmatsu is a member of Deloitte Asia Pacific Limited and the Deloitte organisation. For more information, please visit our web site at https://www2.deloitte.com/au/en.html.Deloitte Asia Pacific Limited is a company limited by guarantee and a member firm of DTTL. Members of Deloitte Asia Pacific Limited and their related entities, each of which are separate and independent legal entities, provide services from more than 100 cities across the region, including Auckland, Bangkok, Beijing, Bengaluru, Hanoi, Hong Kong, Jakarta, Kuala Lumpur, Manila, Melbourne, Mumbai, New Delhi, Osaka, Seoul, Shanghai, Singapore, Sydney, Taipei and Tokyo. Liability limited by a scheme approved under Professional Standards Legislation.© 2026 Deloitte Touche Tohmatsu.  More CoverageWhat kind of AI nation does Australia want to become?JOANA VALENTE and STU SCOTISTech workforce is the key to our nation’s AI futureJOHN O’MAHONY and NICK HULLAI infrastructure: why private options are gaining groundSteve Hebert, Matt Jacobs and Nicholas Merizzi Read related topics:Artificial Intelligence