Challenges and gains of AI in real-life scenarios in the financial sector
This week we participated in one of the ‘The future of AI and Automation in Financial Services’ workshops organized by the World Economic Forum in Zurich. The goal of these workshops, that are currently being held around the world, is to understand the challenges and benefits that Artificial Intelligence (AI) can bring to the table by talking to those that have been in the front line of the finance-AI interaction: FinTech CEOs, Chief Data Scientists in established banks, or Professors in the field. The conclusions will be presented in Davos to inform and influence high-ranking decision-makers from the private and public sectors.
Here, we give a small summary of the main points that were discussed and that seemed most interesting.
As all actors in the field know, using AI is not an out-of-the-box feature that can be easily deployed. There are many challenges ahead before fully enjoying the benefits brought by such an innovative technology:
We need to learn to walk before we can run. Many of the challenges of deploying AI-based applications currently are related to the lack of data integration and normalization within banks. In addition, one often needs to deal with outdated platforms. The cost in work hours and software can push decision-makers to rethink this process.
Technology is ready, culture may not be. Although, the AI algorithms can be very effective, there still is a need for users to understand how to work with them and what their liabilities are. Thus, education is a key aspect of AI implementation.
Sharing. A big issue that is dragging innovation in the sector is the unwillingness to share data or models between the financial actors, mostly due to privacy issues.
All the workshop-participants were AI enthusiasts, so the easiest point to address was the benefits of using AI within Financial institutions.
Financial inclusion and client gain. KYC procedures right now are being over-secured because financial institutions do not trust 100% their risk assessment algorithms and therefore, prefer to be conservative regarding their clients’ onboarding. Thus, having better systems would allow to onboard more people that are currently rejected.
Cost savings. Once the system is well integrated, many procedures can be automated, such as accounting declaration or regulatory reporting. Moreover, fines can be avoided since many regulatory controls can also be automated.
More effective and efficient procedures. The system can be more effective in time and resources, and also provide a better visibility of the challenges of the organization by reporting and summarizing the breaches in its procedures.
The roles of the incumbents, FinTechs and large technological corporations, are different but can coexist. Although big tech corporations can afford years of research and development to obtain cutting-edge technology, they do not have the flexibility to react to the demands of the sector (for instance IBM Watson). Here is where FinTech can work, by focusing on a niche application of current technology to cover specific needs. Big tech companies are also moving towards the banking sector (Google bank?) but this will not happen any time soon.
During the discussions, there was a complete rejection of the idea of centralized institutions taking over the control of certain procedures (AML procedures, for instance). Mostly because the financial institutions are not willing to share their data or models. It was also pointed out that more clear directives and data providing from the regulatory agents would help the field move forward. A clear example is how social network information can be used for KYC procedures.
The AI fantasy for human replacement will not happen any time soon
Although AI can be a game changer in some areas of the Financial sector, in reality, it could only be applied to very few procedures (2% in total was the number pointed out). This was also mentioned in the last issue of Nature ‘Future of work’. Most of the jobs that we think can be completely substituted by IA, in fact, have bigger social communication and decision-making components than we think, which cannot be solved by algorithms. So the most probable scenario will be AI working as a powerful tool so that people can avoid time-consuming mechanical tasks (such as copying and pasting from excel file to excel file all day long).
To conclude, participating in this workshop was very interesting. It gave us great insights into current concerns of Financial actors of all types and geographical areas. Moreover, it was exciting to collaborate on such an important initiative, which we hope will help the sector move forward by bringing clarity and eliminating the gap between the different actors in the field.