I want to walk you through a major step the Commonwealth Bank of Australia has taken to embed artificial intelligence into the way it works and serves customers. The bank is rolling out ChatGPT Enterprise across nearly 50,000 people, and that move is about more than introducing a tool. It is a deliberate effort to build AI fluency, accelerate product delivery, and improve customer outcomes across an organisation that serves more than 15 million Australians.
🔎 What’s happening at CBA
The bank has entered a deep partnership with OpenAI to deploy ChatGPT Enterprise widely across the business. This is not a pilot with a handful of teams. It is a broad-scale rollout intended to reach tens of thousands of employees and touch many parts of the customer journey—from account opening to high-anxiety moments like suspected fraudulent transactions.
As Matt Comyn, CEO of the Commonwealth Bank, put it:
"We're really focused on making sure we're delivering a better customer experience across the whole organisation and working closely with OpenAI to roll out ChatGPT Enterprise across almost 50,000 people."
That quote captures two clear priorities: customer experience and scale. The bank isn’t just experimenting with AI; it is trying to create consistent improvements that customers can feel across different products and touch points.
🎯 Why this matters for customers and the broader economy
The Commonwealth Bank is Australia's largest financial institution, and its decisions ripple across the economy. When a bank that serves millions adopts AI at scale, the potential impact is threefold:
- Faster, smoother customer experiences. Routine tasks like account opening, loan origination, or inquiries about suspicious activity can be streamlined and handled more empathetically and accurately.
- Improved fraud detection and response. AI can surface patterns and automate responses to scams and suspicious transactions, reducing financial harm to customers.
- Faster product development and delivery. Teams equipped with AI tools can prototype, iterate, and bring new services to market more quickly, which benefits customers and supports economic growth.
Matt highlighted the real-world customer moments they want to improve: a nervous customer dealing with a suspicious card transaction, or someone navigating the anxiety of applying for a loan. Those are the interactions where timely, empathetic support matters most.
🚀 How CBA is approaching the rollout
Rolling out an enterprise-grade AI tool across tens of thousands of employees requires a plan that goes beyond software deployment. From what I see in the bank’s approach, the crucial components include leadership modelling, broad education, targeted use cases, tight partnership with the platform provider, and governance.
Leadership as role modelling
Leadership behaviour matters. Matt Comyn was an early adopter of ChatGPT personally, and that kind of senior buy-in is invaluable. When senior leaders use a tool and communicate its value, it reduces resistance and encourages adoption.
The bank held a leadership forum where a significant portion of the agenda focused on OpenAI and ChatGPT. That signals to the organisation that this is strategic, not experimental.
Scale through familiar tooling
A key decision was to use tooling people already understand. Deploying a solution that feels familiar lowers friction and helps employees start delivering value faster. That means less time on basic training and more time applying the tool to real problems.
Partnering deeply with the platform
CBA’s approach is explicitly collaborative with OpenAI. A deep partnership helps the bank align the technology with regulatory, security, and product requirements. It also makes it easier to tailor the platform to the needs of bankers and customers while maintaining enterprise-grade controls.
Targeted, high-value use cases
The bank is prioritising use cases where AI can deliver measurable improvement quickly:
- Account opening and onboarding
- Loan origination and documentation
- Customer support interactions, including fraud and scam responses
- Internal productivity tasks that free up time for more strategic work
Targeting these high-impact moments helps deliver visible wins that justify further investment and rollout.
🛡️ Fraud, scams, and financial crime: a top priority
One use case Matt emphasised is fraud and scams—an area described as "near and dear to our heart." Financial institutions are on the front line protecting customers from increasingly sophisticated attacks. AI has a potent role to play in both detection and response.
The bank is thinking about how AI can:
- Detect unusual patterns in real time.
- Provide frontline staff with better explanations and recommended actions when they encounter suspicious activity.
- Help automate communications to customers to reduce the time between detection and mitigation.
Using AI for fraud and scams is not just a technical challenge. It is an operational and ethical one. Rapid detection must be balanced with clear customer communication and safeguards against false positives. When properly implemented, AI can reduce financial harm and increase public trust in the financial system.
📈 What success looks like: measures and outcomes
To know if an enterprise AI rollout is working, organisations need to track outcomes. Based on CBA’s stated goals and common practice, these would include:
- Customer experience metrics: Net Promoter Score, customer effort scores, complaint volumes for targeted journeys.
- Operational efficiency: Reduction in manual handling time, faster processing times for applications, and resolution times for fraud investigations.
- Adoption and fluency: Number of employees actively using the tool, training completion rates, and satisfaction among staff.
- Risk and compliance measures: Incidence of misclassification, false positives in fraud detection, and adherence to data governance policies.
- Product acceleration: Time-to-market for new features or products developed with AI assistance.
Clear KPIs let teams iterate on deployment, scale what works, and course-correct where outcomes fall short.
🧭 Governance, privacy, and responsible deployment
A large bank cannot simply flip a switch and hand out an AI tool. Data governance, privacy, security, and regulatory compliance must guide every step. While Matt did not detail CBA’s governance framework in the remarks I’m referencing, a responsible rollout usually includes these elements:
- Data classification: Define which data can be processed by AI models and which cannot. Protect customer identifiers and sensitive financial information.
- Access controls: Role-based permissions to limit who can use AI features and what data they can access.
- Audit trails: Logs of prompts, model outputs, and decisions taken based on AI recommendations to enable reviews and compliance checks.
- Human oversight: Clear rules for when humans must review or override model outputs, especially in high-stakes decision points like credit approvals or fraud lockouts.
- Model evaluation: Ongoing testing for bias, accuracy, and robustness, with plans to retrain or tune as needed.
- Vendor collaboration: Contractual guarantees around security, data handling, and support from the AI provider.
These guardrails protect customers and the bank’s reputation while enabling teams to innovate safely.
🧠 Building AI fluency across an organisation
Delivering a tool is not the same as building fluency. Fluency means employees understand what AI can and cannot do, how to use it responsibly, and where it adds the most value.
From what I observe in CBA’s approach, the bank is taking a mix of top-down and bottom-up actions:
- Leadership modelling: Senior leaders using the tool and talking about it openly to signal strategic priority.
- Practical training: Role-specific training that shows how the tool changes particular workflows, not just general hype sessions.
- Communities of practice: Cross-functional groups that share successful templates, prompts, and guardrails.
- Internal champions: Early adopters who help their teams integrate AI into daily work and mentor others.
- Measurement and feedback loops: Collecting usage data, success stories, and areas of friction to refine deployments.
Fluency also requires cultural changes. Teams need permission to experiment, fail fast, and share learnings. In a regulated industry like banking, that culture must be paired with disciplined governance.
📚 Practical examples: where AI already helps
There are concrete moments where generative AI can deliver immediate benefit. I’ll sketch several scenarios that match CBA’s priorities and that other organisations can emulate.
Account opening and onboarding
Onboarding often involves forms, document checks, identity verification, and explaining terms to customers. AI can:
- Auto-fill and validate application fields to reduce friction.
- Summarise product terms in plain language, improving transparency.
- Provide contextual help during the process to reduce abandonment.
Loan origination
Loan origination involves gathering documents, assessing eligibility, and producing loan offers. AI can accelerate document classification, highlight key risk factors, and generate initial offer drafts for human review.
Customer support and empathy at scale
Customer-facing agents and contact centres benefit from AI in two ways:
- Pre-call support: Summaries of recent customer interactions, account status, and suggested scripts to reduce resolution time.
- Post-call actioning: Auto-generation of follow-up emails or next-step instructions to ensure customers know what to expect.
Responding to suspicious transactions
In the high-anxiety moment when a customer receives notification of a suspicious transaction, response speed and clarity matter. AI can:
- Aggregate contextual information quickly for the agent.
- Provide a recommended course of action based on rules and past outcomes.
- Draft clear communications for customers that reduce confusion.
These capabilities create relief for customers and reduce the mental load on staff handling stressful interactions.
💡 Lessons for other organisations
Whether you work in banking, insurance, health care, or retail, CBA’s approach offers several transferable lessons:
- Start with leadership alignment. When senior leaders adopt and advocate for AI, it creates momentum.
- Pick high-impact, bounded use cases. Focus on customer moments where AI can demonstrably reduce friction or harm.
- Use familiar tooling where possible. Familiarity reduces adoption friction and accelerates impact.
- Partner with your technology provider. A collaborative vendor relationship helps tailor the platform to your compliance, security, and operational needs.
- Invest in governance early. Build guardrails, auditability, and human-in-the-loop processes before scaling widely.
- Measure outcomes, not just usage. Track customer metrics, operational improvements, and risk indicators.
- Build communities and training paths. Move from the exploratory to the practical with role-focused education and peer learning.
These steps help organisations capture value while managing the inevitable risks of deploying powerful new capabilities.
🔁 Building momentum: product acceleration and iteration
One of the outcomes Matt mentioned is the ability to bring products to market faster. I want to expand on why that happens and how teams should manage it.
AI accelerates parts of the product development lifecycle:
- Ideation and design: AI can generate customer research summaries, draft user flows, and propose hypotheses to test.
- Prototyping: Teams can produce mock-ups, content, and scripts faster.
- Testing and iteration: AI can help analyse feedback and suggest prioritisation for fixes.
Faster does not mean ungoverned. I recommend these practices when using AI to accelerate product work:
- Keep humans in the loop: AI should augment decision-making, not replace it, especially for regulated products.
- Define acceptance criteria: Ensure quality gates remain strict even if iteration is faster.
- Track experiments: Maintain an experiment registry to ensure learnings are captured and reproducible.
⚖️ Balancing innovation with accountability
The tension between speed and accountability is real. Many organisations rush to unlock productivity gains but neglect the processes that keep customers safe. I believe the best practice is to treat responsible deployment as a competitive advantage: it enables broader trust, higher adoption, and fewer costly remediations down the line.
Practical accountability measures include:
- Pre-deployment risk assessments that evaluate potential harms and mitigation plans.
- Post-deployment monitoring for unexpected behaviour, drift, and customer feedback.
- Clear escalation paths when AI outputs are disputed or appear incorrect.
- Regular audits for fairness and model performance across customer segments.
🛠️ A practical checklist for organisations starting a similar journey
If you’re thinking about scaling an enterprise AI tool, here’s a concise checklist based on what I’ve learned from CBA’s example and industry best practices:
- Leadership alignment: Secure visible sponsor(s) and include AI strategy in leadership forums.
- Use-case prioritisation: Rank potential use cases by customer impact, regulatory risk, and feasibility.
- Vendor partnership: Establish a technical and legal framework with your provider for security, data handling, and support.
- Governance framework: Define data policies, access controls, human-in-the-loop requirements, and audit processes.
- Training and change management: Build role-specific pathways and communities of practice.
- Monitoring and KPIs: Set clear metrics for customer experience, operational efficiency, and risk indicators.
- Pilot and scale: Start with a controlled rollout, collect evidence, then expand to broader teams.
- Continuous improvement: Regularly review model performance, user feedback, and compliance requirements.
📣 What to watch for as this evolves
As the bank continues to scale AI, several indicators will tell me whether the effort is having its intended effect:
- Consistent customer improvements: Fewer complaints, faster resolution times, and higher satisfaction scores across targeted journeys.
- Operational lift: Clear reductions in manual processing time and visible productivity improvements for frontline and back-office staff.
- Reduced financial crime impact: Fewer successful scams, faster remediation, and fewer customer losses attributable to fraud.
- Healthy adoption metrics: Sustained usage among employees, not just spikes during initial rollout.
- Strong governance outcomes: Demonstrable audits and review cycles that keep risk in check.
🧩 Final thoughts
The Commonwealth Bank’s initiative is a useful case study in how large organisations can move beyond pilots to truly embed AI into everyday work. The combination of leadership buy-in, partnership with the platform provider, focus on high-impact customer moments, and attention to governance is the pattern I would recommend to any enterprise.
AI is not a silver bullet, but when applied thoughtfully it can reduce friction, speed up product delivery, and protect customers from harm. The work CBA is doing—bringing ChatGPT Enterprise to tens of thousands of people with a focus on fraud prevention and customer experience—shows how scale and responsibility can go hand in hand.
I’ll be watching how customer metrics and financial crime outcomes evolve as the rollout continues. For now, the bank’s approach is a reminder that technology matters most when it is paired with clear leadership, practical governance, and measurable goals.
