Introduction

With OpenAI’s ChatGPT gaining a 100 million users in first 2 months of launch, Generative AI has become the latest buzz word to dominate the corporate circles.

Google soon followed ChatGPT with Google Bard, and Microsoft has already announced the “future of search” by integrating GPT into Bing and Edge.

In simplest terms, Generative AI is a system with artificial intelligence that can be used to generate content (text, images, videos, audio, code, synthetic data, …). It is by no means new or sudden – it has been used for a while to enhance images and audio quality.

In essence, AI is a set of algorithms that enable an analysis of a large amount of data and interpret and respond to an incoming query. GPT is Generative Pre-Trained Transformer, like OpenAI’s GPT-3, or Google’s LaMDa. ChatGPT as an example has been trained on 45 Terabytes of data, equivalent to millions of volumes of books. Both ChatGPT and Google Bard can interpret questions asked in common language and respond with relevant information. In addition, they use Generative Adversarial Networks (GAN) to enhance / generate new information.

The technology has opened a lot of discussions on its use cases. While it is still early times, I am sharing my perspectives on what it could mean and how the sectors could start thinking about and prioritizing their investments in this space.

Potential applications to the financial services industries

I see primarily 3 main (potential) use cases in the banking, insurance and payments sectors.

  1. Fraud prevention: Generative AI can be used to generate large volumes of test data that caters to different combinations otherwise difficult to generate. This provides a much more robust testing strategy for products and enable tightening of rules to prevent fraud and flag anomalies. In addition, the generated test data can also be used to train an AI ML (machine learning) based fraud prevention module.
  2. Document processing: Financial institutions deal with a large number of documents, many for regulatory reasons. This technology can help automate document processing and reduce the number of errors. This also frees up employee time so they can be redeployed in more value creation. Finally, automated document processing means smaller lead times for communication, and hence improved customer experience. This can further lead to a better NPS and increased revenue.
  3. Onboarding/ loans: Generative AI also enhances the ability to process images and so can make the KYC process much smoother compared to today. A quick but robust KYC implies lower risk for the bank while offering an onboarding / loan decision within minutes, hence improving conversion rates (by reducing the chances of customers dropping out of the process) while reducing the human effort required in the process.

In addition to these 3 use cases, there are also 2 supplemental uses cases, by using Generative AI as an enhanced chat bot.

  1. Customer support: Chat bots are not new and have been discussed or used for over 5 years. IVRs are even older. In both systems (Chat bots/ IVRs), the user is guided through a decision tree to reach the right answer or get human assistance. With Generative AI, the quality of answers can be significantly enhanced. In addition, the number of steps required to reach the answer can also be reduced. This reduces the cost of customer support while providing an improved customer experience. A much-discussed example is offering financial planning advice to customers by analyzing their information, but the regulatory and privacy aspects of this use case remain unclear at the time of writing of this article.
  2. Knowledge management/ Staff training: Generative AI can also assist internal staff in finding information, addressing customer queries more efficiently, solving problems by accessing known solutions faster and in their onboarding or training. This can significantly improve the skill level of the employees to ensure consistent and high-quality customer service. This can also enable employees to upsell or cross-sell better leading to higher revenues. Finally, it should lead to higher retention rates in employees hence reducing the cost of HR processes (including hiring and exits).

Finally, there is also a use case as an enhanced research tool

  1. Market trends/ investment support: Generative AI can potentially also be used to interpret the impact of events around the world on potential investment decisions and hence help shape the risk profiles accordingly. The main value here will be in joining the dots in existing search results and providing relevant insights.

Risks and open questions

As with every young technology, there are several open questions that need to be addressed before the solutions could be industrialized. These range from regulatory (how do we ensure a fair use of data, how do laws like GDPR apply), confidentiality (how will the data be stored and used, will AI providers / competitors be able to see it), ethics and privacy. Does Generative AI offer sources to citations or will it encourage plagiarism?

Additionally, it requires a huge amount of computing power and data to run a system like this and be effective. How feasible will it be for “normal” companies to be able to run this, or will they be at a significant disadvantage vis-à-vis larger companies that can? Do companies use a common hosted service, and what happens if that service runs into issues or goes down?

There is some work to be done before this becomes mainstream, but needless to say, Generative AI promises a plethora of exciting innovations and disruption in the times to come.