The personal computer and the internet have had a vast socio-economic impact in the 20th century. Where computers have (partially) automated many administrative and repetitive jobs in the 20th century, Artificial Intelligence (AI) provides the key to unlocking the automation of less rule-based and more cognitive oriented jobs in the 21st century.
AI is considered by many as the key driver to productivity growth in the 21st century. Despite the hype there are, in my opinion, five key considerations before investing in AI. From my experience, these considerations have a variable impact depending on the size, shape and industry. This article displays how failing to properly examine a number of aspects will lead to less than optimal implementations of the technology. These considerations should not be seen as hurdles to the success of artificial intelligence, but as a checklist to test your organizations’ readiness to make optimal use of it.
Ensure That Your AI Will Be a Success
The technology behind AI is complex and its success depends on the environment in which it is implemented. There are a number of preconditions such as the quality of your data, but also your clients readiness to engage. Furthermore, it is crucial to keep in mind that AI has significant impact on the transparency of your decisions and your talent management procedures. Below I will discuss each of these in detail.
1. The Success Stands or Falls Based on the Quality of Your Data
For artificial intelligence to work well, you need to have relevant, accurate and well-structured data. As highlighted in McKinsey Quarterly, the power of data is often locked into fragmented and siloed (legacy) systems. Over the years, many financial institutions have seen their data and systems architecture become ‘spaghetti’ like. However, I cannot stress this enough: the data is just as important as the AI technology itself, since AI builds upon the data. If data is not correct, precise or relevant, then the AI will make decisions which may not be the most accurate one. Consider either long-term structural golden data sources and governance changes or short-term patches based on your strategy and roadmap.
2. Decisions Made by AI Are Not Easily Understood
You need to realize that artificial intelligence is not explanatory. Meaning that unless you are an expert data-scientist, you will not be able to find out why your AI came to a certain conclusion. Although a lot is being invested by organizations, such as DARPA and companies like Accenture, ‘Un-explainable AI’ could potentially become a problem. The upcoming Global Data Protection Regulation (GDPR), by which consumers shall have “the right to receive a justification of the automated decision” as summarized by the Information Commissioner’s Office, may further complicate matters.
3. Not All of Your Clients Are Fully Ready to Deal With an AI (yet)
Although a recent Accenture report revealed that 70% of consumers would welcome an AI-advisor for their banking, insurance or retirement related services, you need to accept that not all are ready to interact with artificial intelligence yet. Some customers may feel their privacy is being infringed upon, while others just wish to (incidentally) speak with a human. This will depend in large extent to your target audience’s demographics and (previous) experiences.
It is best to be transparent and open to the client about the benefits of AI, in order to ensure that clients who do dip their toes into the water are not led into abandonment. In other words, ensure that your clients who are willing to interact with your AI leave the interaction satisfied. This can be done through pre-emptively responding to their queries. Your AI can monitor and analyze an almost infinite amount of data sources such as in-app/website queues to detect distress. You will need to train your AI to react instantaneously by offering support through FAQ’s or personalized help.
4. Ensuring Your AI Remains Ethical
Another interesting topic in the debate of AI is ethics. In general, the list is long when it comes to ethic debates in the context of AI. Take for example racial biases: this machine was biased towards African-Americans when it came to predicting crime. How would a similar instance of AI technology handle racial differences when considering a mortgage application? In some cases, it may be fair to point out that machines learn from human mistakes. In other cases, however, AI can be subject to malicious intervention as highlighted by the Harvard Business Review. They point out that although artificial intelligence is expected to play a pivotal role within security, ensuring that an AI is not either hacked nor programmed with malice may prove more important and difficult than one may expect.
Within financial services, I believe that the most important point of discussion is bias and privacy. More specifically, if an AI has access to an incredible pool of data stored within the bank or insurer but also has access to social media and other external databases, how do we ensure a bias does not present itself? Or safeguard the privacy of all stakeholders involved? There are a number of additional matters to consider, such as safeguards and accountability. These are outside the scope of this article.
5. Alter the Way You Acquire, Manage and Retain Talent
Talent management and acquisition will also be impacted. It is crucial to communicate clear expectations to your people, invest in continual learning to avoid a mismatch in skills, and realize that you may need to look externally for certain specialists (e.g. data scientists, deep learning specialists etc.). Not only does your existing workforce need to remain aware of the developing requirements for their positions, they may also need to be retrained. Similarly, upon adopting certain AI technologies, you will require project managers, business analysts, architects and specialists with experience in the AI domain to maintain and improve you AI operations.
Recapping the Two Sides of the AI Coin
As we have seen in the article How artificial intelligence helps banks and insurers achieve their strategic priorities, employing AI within your organization can go a long way towards achieving your strategic priorities for 2017. AI has the potential to:
- Provide customers with higher levels of service availability and personalized experience;
- Increase regulatory compliance through more (cost) effective control procedures (e.g. KYC);
- Cut costs required to run the business and identify new sales;
- Increase the satisfaction of employees by removing mundane tasks from their work;
- Extract additional value from the mountain of data stored by the bank or insurer.
However, as we have mentioned, there are a number of considerations to keep in mind, such as the quality of your organization's data, the (un)explainable nature and the readiness of your organization and its processes. Furthermore, you will also need to tackle key questions around ethics and talent management.
You Want to Get Started With AI. What’s Next?
Just like the computer excelled global productivity growth in the 20th century, AI has the potential to unlock the next advancement in the 21st century. Even though we are already witnessing human resistance. Looking at technological advancements over the past, it is always better to be ahead than to get caught staring.
The first step is to acquire sponsorship at board level. A multi-disciplinary task force and sufficient budget should be made available to explore the possibilities and identify opportunities within the organization. Starting small, focusing on quick wins and experimenting a little, is a good way to create momentum and generate followers in your path to intelligent automation. Furthermore, if your bank or insurance company is in the middle of a digital transformation, as most are, its crucial to align your artificial intelligence roadmap to the objectives of your (new) digital business model.
Although AI, in general, has not reached maturity levels of technologies such as robotic process automation, it is picking up the pace. We have looked at the possible benefits to banks and insurers in achieving their strategic priorities for 2017. We have seen that the benefits of artificial intelligence are wide and diverse. This is no different for risks. Banks and insurers who manage to embed AI into their current organization will see growth opportunities that are not seen in decades. Banks have weathered storms, but in 2017 there should be no more excuses. The time has come where banks and insurers no longer chase the facts, but instead set the standard in service and profitability across all industries worldwide.