ANZ Bank – Mercury Books | Digital business books and other titles

Gregory John Lee

Vernon Timothy Thaver

From

Robotic process automation at ANZ Bank

Main area: Technology in business (automation)
Other possible areas: Business cases for digital transformation

In a worldwide financial sector that has been deeply disrupted by technological advancements and opportunities, Robotic Process Automation (RPA) stands as one of the most transformative technologies making waves in this industry. ANZ Bank, a leading Australian-based financial institution, has been at the forefront of adopting and implementing RPA to streamline its operations, enhance customer experience, and improve workplaces for its employees.

Introduction to ANZ Bank

ANZ Bank is headquartered in Melbourne, Australia. It has grown to become a multinational financial institution, with millions of customers in some 30 countries across various regions notably in the Oceania and Asian territories (ANZ Bank, 2023b). It consistently tops ratings in a variety of financial services categories in the region (ANZ Bank, 2023a).

ANZ Gets into Automation

Recognizing the potential advantages of Robotic Process Automation, the bank was an early adopter of widespread RPA, beginning its journey in 2015 (SSON Network, 2015). They embarked on a comprehensive journey to integrate automation into various areas, with multiple aims such as saving costs, improving operational efficiency, meeting regulatory requirements more accurately, and so forth. Areas into which ANZ Bank has strategically deployed RPA include back-office operations, customer service, employee processes, compliance, and risk management (Crozier, 2017; Intelligent Automation Network, 2017).

Early Benefits

As reported by SSON Network (2015), implementation was rapid and, even in the early days of the project, advantages were quickly apparent:

ANZ started on a Robotic Process Automation (RPA) journey in early 2015. Beginning the deployment in its Bengaluru global in-house center, it quickly ramped up across all four of its Global Hubs in Asia and Pacific. The program rapidly scaled to over 100 robots, with another 100 expected in the coming quarter, and nearly one thousand more in 2016. After witnessing the strong start in the Global Hubs, ANZ businesses in various countries have also begun to adopt RPA, leading to this initiative being one of the fastest growing automation programs around … the results to date [showed] ROI within one year. With 10,000 people in ANZ’s 4 Asian delivery hubs, the positive impact of RPA is enormous – and not just in terms of improved reliability and quality. Today, staff are able to spend more time analysing what the data means, rather than try to make the process work. And that is the kind of value-add that drives success.

Workforce Savings for ANZ

By automating or semi-automating repetitive, rule-based tasks, RPA has allowed the bank to free up valuable human resources for more strategic, value-added activities. A key part of the focus in ANZ’s Robotic Process Automation was to bring staff along on the journey, making them feel included and served by the automation process rather than fearing impacts on their jobs. As stated by Simen Munter, general manager of ANZ Group Hubs in 2017 (Intelligent Automation Network, 2017):

…it’s people, process, technology, in that order. So what we want to do is to do things with our people. And that’s a key element for me in this. We are not doing this to our people, we are doing it with our people.

Automation for Customer Experience

One of the standout benefits of ANZ Bank’s RPA implementation has been improvements in customer experience. By automating routine processes, the bank has been able to reduce processing times, leading to faster response times for customer inquiries and requests and translating into higher customer satisfaction and loyalty (Crozier, 2017; Intelligent Automation Network, 2017).

Operational Efficiencies

RPA has also played a key role in streamlining ANZ Bank’s operations. Tasks such as data entry, reconciliation, and transaction processing, which were previously time-consuming and error-prone, are now executed with precision and speed by software robots. This has resulted in improved operational efficiency and a significant reduction in error rates (SSON Network, 2015).

Compliance Benefits for ANZ

In the highly regulated financial industry, compliance with regulatory standards is paramount. ANZ Bank has leveraged RPA to enhance its compliance efforts. Software robots are programmed to monitor transactions and processes for compliance with relevant regulations and to identify and rectify any discrepancies in real-time (Dalton & Spiteri, 2021). This proactive approach to compliance has not only reduced the risk of penalties but has also instilled confidence in regulators and customers alike.

Sustained Improvements at ANZ

These improvements were most visible going into the 2020s after the first few years of automation, and before COVID-19. During this period, employee engagement increased by over 10%, retail customer net promoter scores achieve all-time highs, and operational criteria such as customer complain resolution improved (ANZ Bank, 2020).
By automating routine tasks, ANZ Bank has achieved substantial cost savings (Dalton & Spiteri, 2021). The reduction in manual intervention has led to lower operational costs and a more efficient allocation of resources. This has allowed the bank to redirect resources towards strategic initiatives, such as innovation, product development, and customer-centric initiatives.

Automation and Competitive Strategy at ANZ

One interesting feature of the bank’s automation drive is its position in relation to competitive strategy. In an industry that is increasingly competitive with the introduction of new, lightweight fintech players, ANZ has had to wrestle with how it can compete as a large, traditional player with slow, hard-to-adapt, legacy IT systems. As related by Dalton & Spiteri (2021):

Traditional banks like ANZ face three broad choices if they are to hold onto their customers and drive longer term value propositions. They can adopt a mindset of continuous improvement of existing platforms, products and services – minimising disruption but ultimately leaving themselves vulnerable to the new competitors. They can start anew, building (or perhaps buying) a modern, digital bank from scratch, leaving legacy challenges behind in an “old” bank which is eventually allowed to wither. Or they can modernise from within, simplifying platforms and offerings, dramatically improving customer experience, creating a fundamentally different customer proposition on modern systems that are both digital and in-person.</br />
There are obviously pros and cons for all approaches but having now spent some years analysing these trends globally and better understanding our own business, it is this third approach we have adopted at ANZ.

To achieve this aim, ANZ launched a parallel brand called ANZ Plus. ANZ Plus focuses on product simplicity with great customer experience, and aims to improve customer financial sustainability (Dalton & Spiteri, 2021). The brand is built on modern, digital platforms and heavily based on self-service backed by extensive automation, while still including some elements of direct customer service. RPA will, therefore, be a core element of the future strategic direction for ANZ Bank.

Challenges for ANZ

While ANZ Bank’s foray into RPA has been largely successful, it is important to acknowledge that the implementation of such transformative technology is not without its challenges.

Change management can be a daunting task when automation is involved, especially as workforces remain suspicious of such initiatives notwithstanding the focus on bringing people with them that ANZ highlighted earlier in the case study (LIGS University, 2024).

Integration with existing systems is sometimes tricky for RPA (TecQubes Technologies, 2024), which is why they made the choices with ANZ Plus mentioned earlier, although legacy IT continues to be a challenge for large banks.

Finally, RPA requires increasing amounts of governance and maintenance, with a growing tram required to monitor and look after existing robot workforces while expanding the portfolio (Boulton & Olavsrud, 2024).

Conclusion: ANZ and Automation

Looking ahead, ANZ Bank is poised to continue its journey with RPA, exploring advanced applications such as cognitive automation and artificial intelligence. This will further enhance the bank’s capabilities in areas like data analytics, fraud detection, and personalized customer experiences.

ANZ Bank’s strategic adoption of Robotic Process Automation has exemplified how technology can be harnessed to revolutionize traditional banking operations. The integration of RPA has not only resulted in operational efficiencies but has also elevated the customer experience and bolstered compliance efforts. As ANZ Bank continues to explore the frontiers of automation, it stands as a beacon for the financial industry, demonstrating the potential of RPA in shaping the future of banking.

One thing seems certain: having had a taste of the power of RPA to streamline and automate processes, ANZ will never go back.

Challenge to the Reader: ANZ Bank

Imagine you are a consultant who has been approached by ANZ and given the information embedded in this case study and no other information. The bank wishes to know what avenues you might suggest to help them consolidate and expand their automation journey in the future. What would you tell them?

Reflection on the Challenge: ANZ Bank

As a consultant who has been approached by ANZ to suggest avenues to consolidate and expand on their automation journey in the future, you could consider the following:

Creating a dedicated cross-disciplinary “Automation Center of Excellence”.

The information in the case seems to indicate that initial implementation, at least, was relatively decentralized and perhaps even haphazard. The bank could benefit from a fully-fledged automation center of excellence (CoE). Such a CoE would have various roles as detailed in Lee & Armstrong (2023), including:

Management of the CoE,
Bot developers and other technical teams,
Staff responsible for managing the existing portfolio of projects,
Staff who connect the CoE to the company’s business units (Lee & Armstrong, 2023, argue that these would optimally be embedded in the business units as business partners, but reporting to the CoE), and
Change management experts, responsible for implementing the changes once the bots are developed.

Note that it may be likely that, in reality, ANZ has implemented an automation COE. This is not mentioned in the case study or publicly available documents. However, as a consultant, you would want to find out and audit whether the CoE has been designed adequately with respect to all possible capabilities.

Another critical point is that automation CoEs of this sort have traditionally been focused on RPA only. However, as detailed in Lee & Armstrong (2023), the wider variety of automation types have become inextricably interlinked in the “hyper-automation” concept. The next bullet point discusses including more AI into the ANZ automation portfolio. Other automation angles also apply, such as IoT, physical robots, metaverse components in the future, and so on. This requires a CoE with a broad base of skillsets, certainly broader than the narrow RPA teams that organizations used to put together.

Implementing increasing AI integrations

This case study focuses on traditional RPA. There is little to no artificial intelligence mentioned. Although ANZ has doubtless implemented a range of AI options from packaged options within their subscriptions to models of their own creation, you would want to audit this and determine if the bank is taking full advantage of the current AI explosion. Notably, there are three major arenas for investigation:

Packaged AI ripe for subscription. There are a variety of packaged AI algorithms, some likely within the bank’s current software subscriptions and some open for subscribing to, which might be harnessed for almost immediate gains by folding them into automation. For instance, the Microsoft Power Platform suite, including Power Apps and Power Automate among others, now contains directly embedded AI modules / actions by OpenAI for natural language understanding and machine perception tasks (decoding pictures). If the bank were using the Power Platform, by adding this AI subscription, they can immediately begin including AI tasks in their automations such as language detection, translation, sentiment analysis, language classification (e.g. as complaints, compliments, queries, etc.), summarizing text, analyzing text for entities (e.g., brands or business units), decoding scanned documents (such as IDs, receipts, invoices, and business cards), describing images, and so on. Other RPA suites like UI path and others include many similar options. These tasks encompass a great number of business-relevant requirements. Similarly, there are a great number of stand-alone AI APIs into which the bank can tap. As a consultant, you would want to assess current usage and help the bank align priority subscriptions to its strategy.
Custom AI models built by the bank to automate prediction. The bank may also wish to build custom AI models using its own data to help with tasks that are not adequately achieved by standard out-the-box subscriptions. For instance, one AI task offered by all major language AIs is to extract entities, that is, the people, institutions, or brands about which the text is talking. However, standard / out-the-box AI models may do a poor job in recognizing the bank’s specific business units or brands, because these algorithms are trained on a broad internet base. If these entities are poorly predicted in the language AI, then the overall usability of the AI is compromised. For instance, say the bank wishes to automatically monitor social media feeds for negative-sentiment complaints, and narrow this to what specific business units or brands are being spoken about so that it can a) funnel the complaints to the relevant customer service units, and b) create a dashboard of social media sentiment that is specific to the units or brands. The usefulness of this automation is fundamentally compromised if the AI algorithms cannot, in fact, recognize the business units or brands. Therefore, the bank may wish to train its own custom algorithm to recognize the bank’s specific entities far better than an out-the-box model would. This is highly achievable these days even without serious data scientist inputs (although, you would want to involve such inputs), as the process for training such models has been itself made standard in the cloud, and straightforward by technology companies . As a consultant, you would wish to identify where custom models would add the most value, with the highest correlation to strategic priorities, and help to get these trained.
Generative AI. The previous point applies more to prediction algorithms, which tend to analyze big data like language or pictures to make predictions such as what language it is. Generative AI obviously seeks to do the opposite, i.e., to generate big data from a prompt. There has been an explosion of implementations over the past year (2023 to early 2025 at the time of writing this case). As one example, generative AI can potentially help to generate better outgoing customer communications, for instance, by helping to word excellently written press releases. However, there exists a substantial danger that lazy use of this technology could lead to sloppy or wrong work. For example, in the case of press releases, an AI-generated release should be re-worked by a competent person to ensure that everything in it is accurate, approved in terms of communications policies, and customized to the specifics of the organization since the models used by generative AI produce the most likely thing that would have been produced by their training set (e.g., the internet) and not necessarily the specifically-relevant organizational details (e.g., your pricing). So, as a consultant, identifying the best and most strategically aligned uses for generative AI would be a beginning task, where the bank has not already implemented such technology. Secondly, helping the bank to identify and assimilate the best sources of such AI APIs is important, as this is a rapidly shifting landscape. Thirdly, you should assess whether the bank has adequate governance “guardrails” for the use of generative AI (e.g., in the previous example, rules that outgoing communications that included generated text have to pass certain standards or moderations before release).

As noted previously, it is optimal if both RPA and AI can be governed from the same “Automation Center of Excellence”, which is often not the case. Doing so can help ensure that these two areas of automation, which increasingly work hand-in-hand, are completely aligned. As one further note, you should make sure that as part of the CoE, processes exist for reviewing the AI models, subscriptions and options regularly, given the rapidly shifting landscape in this regard.

Reviewing user interfaces

Ultimately, all automation should interface somehow with the workforce of the bank or its stakeholders. Some automation operates in a behind-the-scenes, unattended manner, such as audit bots which scape through datasets for anomalies, but even these bots should report back to human beings in some way such as reporting those anomalies, building dashboards for human consumption, and so on. Some automations include a human-in-the-look (HITL) component, such as an automated travel expenses claim process that asks for managerial approval for a travel reimbursement. And some automations are triggered by human actions, such as a customer clicking something in an app or speaking to a chatbot. Best practice in such human touchpoints is constantly evolving, and a consultant in this space should review all such interfaces with a view to asking whether there are better ways to design them. For instance, many countries require institutions like ANZ to undergo regular updates on their customers’ identities and details (“know-your-customer” or “KYC” renewal processes). Gathering such information might have been designed using a fairly poorly designed form in an early automation design, including making the customer self-enter information that could be more accurately verified through government datasets. As the consultant, you could help to identify that this is a poor interface and re-engineer it. Similarly, some interfaces work better as apps, some should be switched to or away from chatbots, emails can be better changed to formal approvals (e.g., see Microsoft Approvals), some interfaces should be eliminated and replaced with automatic digital triggers, and so on.

These are only three of a number of constructive interventions that might be considered at this relatively mature phase of a bank’s RPA. We repeat that the actual current activities at ANZ may already involve several of these elements, but as automation becomes more complex constant revisiting of automation will be required.