Home > What’s UP > blog > Three questions banks should ask before implementing AI

blog

07. 13. 2018

Three questions banks should ask before implementing AI

Ever wonder how your bank can benefit from the AI revolution? Here are the key questions you should ask yourself before adopting the new technology, and a few things to watch out for on the way.

Widely expected to bring opportunities to cut costs, pump up revenues and fend off digital disruptors, AI has been long seen as the next big thing for banks. In fact, the financial services sector is among the top three industries, besides healthcare and the automotive sector, where AI is predicted to explode in the near future, according to PwC.

The areas with the biggest AI potential are personalized financial planning, fraud detection and anti-money laundering, back-office process automation and customer-facing operations. And let’s not forget sales and marketing, where McKinsey estimated financial institutions could unlock a potential AI value of up to a staggering $200 billion.

Asking the right questions

There are many examples in and outside the banking industry that show the enormous benefits and endless opportunities AI-driven solutions offer. But how to implement artificial intelligence  and make it work for you? Here are the top questions to think about before buying into the hype.

  1. Are you ready?

First things first, banks should examine if they actually need AI (or maybe there is a non-AI solution that offers a better option) and define the results they want to achieve using AI-powered solutions. According to Florida-based software company Abe, the best way to determine the desired outcome is to balance the cost versus benefit of using AI to handle a certain task.

Next up is building technology teams, or at least establishing a chain of command to manage AI projects, consisting of existing staff members, third-party vendors or both. A specific timeline with clear target dates is also needed to be drawn up for developing, deploying and measuring the effectiveness of the implementation project. Which brings us to:

  1. Develop in-house or seek help?

So, you’re 100% sure that AI fits your bank’s strategy and you’ve set your goals. Great! It’s time to decide whether you want to build AI capabilities in-house or hire an external technology provider. Under Plan A, the most important question is if you have the necessary funds and skills to develop AI solutions and maintain them internally, Infosys says. If you’re buying your AI solution from a third-party provider, however, the first thing you should find out if they have an on-premise or cloud-based delivery model, and whether you are allowed to send data to an external party.

Interestingly, some of the tech giants banks often see as their future competitors have rolled out AI-based services for financial institutions. Most of these focus on user experience and customer engagement: Amazon’s virtual assistant, Alexa, for instance, is already used by JPMorgan Chase, Capital One Financial and other banks in the US.

  1. Do you have the right data?

AI basically lives on data, so you need to have access to the right pool of customer and transactional data before implementing any AI tool. Machine learning solutions, for example, need an enormous amount of information that is standardized, labelled and cleansed of anomalies.

The bad news is that this is all much easier said than done. The good news is that there are more and more vendors who take public sources of data and organize them into data lakes so AI applications can use them. This might be yet another reason to opt for third-party providers, PwC says. And if you’re looking at using AI-enhanced advanced analytics, you’ll also have to make sure to use clean data to avoid potential distortions in results.

Get ready for a few bumps

Banks are already the biggest investors in AI. Infosys’s survey shows that big data automation heads the list of AI priorities, with 65% of organizations having already deployed it or planning to do so. About half of respondents have been eying predictive analytics (54%) and machine learning (51%), while 44% are investing in expert systems and 31% in neural networks. Still, financial services ranks only third from the bottom on Infosys’s AI maturity index.

Not that banks are not eager to harness the benefits of AI. They invest heavily into IT infrastructure (60%), and developing the required knowledge and skills (53%). They also rely on outside help from experts in planning (46%) and knowledge gathering (40%). And what’s in it for them? Most of them are looking into ways to extract value from idle data resources, tackle open banking, get better access to computing power and cloud, and leverage the wide availability of open-source and affordable AI platforms.

But of course, adopting AI is not all unicorns and rainbows. Infosys has pointed out that the biggest barriers include the lack of adequate infrastructure and skills, as well as the lack of knowledge about where AI can be of real help. Other barriers stem from cultural issues including employees’ fear of change, concerns about handing over control, the lack of acceptance and resistance from senior management.

The human side of AI

The relative scarcity of experts and their sky-high costs are another major concern. Bloomberg has reported that even newly-minted PhDs in ML and data science can make more than $300,000 a year. A study by Element AI estimates that 22,000 PhD-level computer scientists around the world have what it takes to build AI systems. But only about 3,000 of them are currently looking for a job, while at least 10,000 related positions are currently unfilled in the US alone. To overcome this problem, big players, like Intel, Facebook or Google, are creating their own AI training programs.

And there are also ethical implications, with the most obvious question being how AI will affect existing jobs. There is no doubt that machines will take over repetitive tasks and routine functions now carried out by humans. But it’s also clear that AI will never replace humans. On the contrary, it is expected to allow people to devote their time to more valuable tasks. After all, AI is not capable of creative thinking, like humans. At least for now.

For more on what banks can learn from other industries and competitors about implementing AI and ML-driven technologies, and how to unlock their power in sales and analytics, download our white paper on AI in banking sales!

AI and ML in digital banking sales