ANZ trials IBM’s Watson in customer service



news Australia and New Zealand Banking Group has revealed it will be one of the first companies globally to trial using IBM’s Watson expert data retrieval platform to attempt to enhance the quality of data available to the bank’s customer service team, in a move that could eventually lead to Watson taking questions from customers themselves.

Watson is essentially an IT platform composed of software running on IBM’s hardware that is capable of answering questions posed in natural language. The platform came to public prominence in 2011 when Watson competed in several public screenings of the US-based Jeopardy television quiz show, eventually winning against a series of human opponents.

However, in a media release issued this week, IBM revealed that it had developed more practical applications for Watson that would usher in what the company described as “a new era of cognitive computer systems”. IBM has developed Watson into a product named the “Watson Engagement Advisor”, a platform which allows Watson to assist customer service agents in organisations with providing data to customers.

“Delivered through cloud-delivered services and online chat sessions, IBM Watson will empower a brand’s customer service agents to provide fast, data-driven answers, or sit directly in the hands of consumers via mobile device,” IBM said in its media release. “In one simple click, the solution’s “Ask Watson” feature will quickly help address customers’ questions, offer feedback to guide their purchase decisions, and troubleshoot their problems.”

In an associated post on an IBM blog, ANZ Bank’s CEO Global Wealth and Group Managing Director, Marketing, Innovation and Digital, Joyce Phillips, noted that the bank was “exploring” the use of Watson within its customer service operations. “It used to be that next generation technology to make humans smarter, faster and at the top of their game was the stuff of cinema and science fiction,” Phillips wrote. “Here at ANZ, we are exploring a groundbreaking solution that holds this very same promise – a cognitive assistant, if you will, that can empower our regional bank advisors to better serve our two million wealth management clients.”

Joyce didn’t describe precisely how the bank would use Watson, or commit ANZ to any specific deployment of the technology, and it remains unclear to what extent ANZ is already or will use the new IBM platform. However, the executive did posit a scenario which may generate some inight into the way the bank is considering the technology.

“Imagine yourself entering a regional bank branch,” Phillips wrote. “You’ve arranged an appointment with the branch’s financial advisor, to discuss the life you want for you and your loved ones upon your retirement, and the solutions you need to achieve it. Two of the key questions you and your advisor need to address are: what types of insurance, retirement and banking solutions do you have now, and do they cover you in a way that will safeguard your assets and yield a secure, well-funded retirement.”

“The fact is, many customers don’t know the answer to that question.”

However, Phillips wrote, financial advisors could have access — assumedly through the use of Watson — to “everything they need to know about a customer’s portfolio, with a click of a button”.

“The natural resource that can fuel a smarter, faster customer experience is Big Data, or the vast amount of electronically generated and stored information – ranging from account information to market research – that can piece together that 360 degree view for the financial advisor. ANZ is collaborating with IBM to give our advisors that ability to harness information and make smarter, faster financial recommendations – enabling a customer experience that is simple, safe and steeped in data-informed insights.”

“The technology that will get us there: Watson, a cognitive computing system that is tailor-made for the customer experience, based on its ability to grasp the subtle context of questions posed in human language, search through a universe of Big Data to find the exact answers its users need, and deliver fast, evidence-based recommendations that can fuel a financial advisor’s advice to clients.”

In a separate statement issued by IBM, the executive added: “We envision Watson to serve as cognitive assistant, adept and quick at making sense of Big Data, that can empower our regional bank advisors to better serve our two million wealth management clients across the region, We are pleased to explore with IBM how Watson can enable smarter, faster financial recommendations – yielding a customer experience that is simple, personalised and steeped in data-informed insights.”

In its statement, IBM said that since its television debut, Watson was “smarter, faster and smaller”. The system has gained a 240 percent improvement in system performance, according to Big Blue, and a reduction in physical requirements by 75 percent. Watson can now be run on a single IBM Power 750 server using Linux — “transitioning from its original size of a master bedroom to that of four pizza boxes”.

I’ve been trying to get my head around this situation with ANZ Bank ‘deploying’ Watson for the past couple of days. Is this, as IBM clearly wants us to think, a radically new artificial intelligence technology which has the potential to overhaul the way large organisations conduct customer service? Or is it just a marketing plot to get a quasi-expert system in to large organisations, coupled with pricey IBM hardware (and, no doubt, consultancy fees to keep Watson ship-shape)?

Essentially … what the hell is Watson? A premonition of true artificial intelligence? Or just a more competent expert system of the type we’ve seen many times before? The conclusion I came up with is that we just don’t know yet.

It’s clear that IBM has done several things very skillfully when it comes to the development and promotion of Watson. Firstly, by giving the damn thing a name and emphasising the use of natural language recognition to interact with it, IBM has created the perception that Watson is not a product, but an entity; not something that you deploy, but something that provides services to you of its own volition. Watson, in the minds of the millions of people who witnessed its performance on Jeopardy, is cute; and a little scary. There must be many out there who wonder whether we’re watching the birth of an evil race of machines such as Skynet from the Terminator series of films, right before our eyes.

This also creates the impression that Watson can learn, grow and develop. If Watson can handle the complexity of taking random questions on a live television show (and win!), then, many of IBM’s customers must have surmised, it should be easy for the platform to handle much more limited datasets, such as those found in banking products and bank customer records.

However, with no real case studies involving the deployment of the technology as yet, it’s unclear as yet as to what extent the new, ‘productised’ Watson will be able to do this, and also to what extent customers and contact centre agents will actually want this.

Consider, for example, the plight of ANZ Bank’s ‘regional bank advisors’, who will doubtless be told by HQ in Melbourne that they are now to seek the advice of Watson whenever possible. Many will, I have no doubt, ignore the advice, believing such an IT system to be irrelevant to their work, or even perhaps offensive. Others will trial Watson, but will, no doubt, find it hard to entirely trust an IT system which they have no real control over. If Watson screws up once, as it inevitably will, many will reject it forevermore.

From an end customer perspective, the potential adoption by ANZ Bank of Watson as a live customer service technology also holds troubling implications. A common frustration of Australians is that as a nation, we dislike dealing with customer service agents in remote locations such as India, who don’t completely understand the nuances of our culture. It’s not hard to imagine that dealing with Watson would be even more frustrating for many customers — kind of like dealing with one of those automated voice telephone systems, but one that perhaps thinks it has self-awareness and free will.

Of course, all of these complaints will quickly fall away if Watson proves itself to be adept at the tasks which ANZ Bank plans to set for it. If the system ‘just works’, and works all the time, then it will be an invaluable asset to the business, and will smooth the path for the adoption of such intelligent agent technology in many other businesses and organisations in Australia and globally.

However, until that evidence arrives, I will continue to be cynical about Watson.

There’s a very good piece of analysis on iTNews this week by seasoned local analyst John Brand, which I think goes into the situation well. Brand argues (and I agree) that the genius here is in how IBM is marketing Watson — not necessarily in the system’s capabilities as demonstrated so far. “Hint: We haven’t seen Watson winning any other game shows yet,” writes the analyst.

In short: If I were a chief information officer closely watching ANZ Bank’s “exploration” of the Watson technology, I would maintain an open mind about the technology. It’s possible, just possible, that Watson will herald the implementation of a whole new category of corporate technology that will solve countless minor problems. In short, a revolution. But it’s also entirely possible that the technology isn’t quite ready for the complexities of real-world usage just yet. I suspect ANZ will find out very shortly.

Image credit: IBM


  1. “eventually winning against a series of computer opponents”

    Didn’t you mean human opponents?

  2. Renai: “Get me that file on Watson actually means, HAL”
    HAL: “I’m afraid I can’t do that, Renai”


  3. ANZ, perhaps more than most of the big four, have problems with having private customer data stored on systems that make it accessible to too many of its 40,000+ internal staff, contractors, service providers and third party vendors.

    Most of their IT spending is directed towards reducing risk. Something like Watson appeals to non-IT literate people controlling the purse-springs because it automagically discovers data (and categorises it into Private, Internal-only and Public information). From what little I know, this is commonly done (Watson may be different) by dynamically designing queries based on what has been indexed, and searches file systems as well.

    IBM would be keen to set it up as the solution to data bloat, to be able to manage the delivery of all data across the bank, and the Bank would like help, as no number of Sharepoint sites can solve the problem of poorly stored and served data sitting in stale in all their filesystems and SQL datastores.

    The risks of doing this are that dynamically mined data will be categorised incorrectly, associations relate to the old systems that the Bank will remove and people forget about. Then there are things the client will not really care about, such as making private data public in the process (though it probably already is) and will be constantly accessed by IBM staff and contractors in offshore jurisdictions.

    In the end the data will be an amorphous mass, poorly normalised, stale and obscure. But it will be in some ways better. This will come down to the amount of work they do to maintain Watson’s performance as it grows fatter and the customisation they and the Bank do around it. Given the laziness of Bank teams to improve or develop anything IT related, it will of course, fail. But IBM will have their data, and they’ll have to pay until they find another quick fix.

    Ultimately, they are just trying to normalise, categorise and clean up the mountain of data they have kept about us over the years, but have failed to manage and maintain. Getting someone else to do it is a great way of avoiding being fired when it goes wrong, costs too much or gets in the way of some of your colleagues’ bonuses.

    • If it looks like a fish and smells like a fish, then maybe it is a fish... just a rotten one If it looks like a fish and smells like a fish, then maybe it is a fish... just a rotten one

      It sound like you are quite cynical about IBM at ANZ? Me too.

      If ANZ has disparate systems with their “Big Data” scattered across these systems, then this is just an attempt by IBM to receive some positive PR and the hands of a non-IT advocate at ANZ. IBM have struggled for some time in Australia to understand the technical requirements needed to consolidate the data into a platform for Big Data.

      IBM’s platforms currently struggle with the load of integrating these systems both on technical and cost levels. Based on the amount of processing required to integrate these systems based on IBM’s SOA platforms its just going to increase the costs significantly with additional load that Watson would require if it was to be successful. And that’s a big “if”. Hence Watson will be very good at only answering questions that have predetermined by ANZ, an more importantly IBM, based on the data they will be able to access. I for one would not to put my financial future in the hands of IBM and Watson.

      Here is a possible scenario;
      Customer: “Watson, what products and services should I consider for my future?”
      Watson: “A new home loan”
      Customer: “Why did you recommend a new home loan, when I already have one?”
      Watson: “Because this is the only data that I am currently connected to at the moment”

      And then there is the problem that most people don’t know the correct question to ask.

      So is it currently its only a win for IBM’s PR team and a passed on cost to ANZ’s customers?

  4. Oh, and I forgot to say they have the added problem of data being stored in all manner of other locations, such as web, application and ‘middle tier’ servers in all kinds of internal, business and production network zones.

Comments are closed.