How to Ignite your Digital Strategy with AI
Be Prepared to Change Mindsets
In 2017, BNP Paribas committed a 3BEUR investment over three years dedicated to the Bank’s digital transformation, creating many new initiatives for the Bank.
Transformation comes with a change of mindsets; it’s by sharing ideas with others and delivering positive results that we can challenge the frontier of things we thought possible.
Increasing Awareness & Engaging the Platform
We often talk about AI, a broad term that often refers to the recent developments of Machine Learning and Deep Learning (DL). DL applications are already everywhere; and their frameworks have proven to be very efficient at multiple tasks: detecting objects on images, translating speech into text, extracting semantic information from text, and many others.
Transformation comes with a change of mindsets; it’s by sharing ideas with others and delivering positive results
At BNP Paribas, we organize conferences and events to raise awareness on stakes, reduce fears, engage our employees, expand their understanding of AI and inspire them to recognize the right use cases. For those most technologically savvy, we regularly arrange hackathons to source ideas and test implementations.
BNP Paribashas a number of digital graduate programs and more than 100 Data Scientist positions open. Even if banks have significant opportunities to apply AI, they are not always the first choice for students when it comes to starting a career in data science. Additionally, implementing AI into our business requires process understanding and an adequate mindset.
To fill the gap, we saw banks recently acquiring AI startups to onboard expertise. In this fast moving industry, BNP Paribas has taken a few minority stakes in startups as a way to enhance partnership with developing companies with whom we can co-create. In 2017, BNP Paribas Securities Services invested in the RegTechFortia, launching a strategic partnership for the custodian business. More recently, BNP Paribas invested in other AI startups such as Digital Reasoning to improve customer insight or Cashforce to offer new services to treasurers.
Be Strategic: Identify, Test, and Implement
AI implementation is not an easy task. Machine learning models are not executed or maintained like traditional IT developments. The expertise is scarce, and there are a number of pitfalls, but the benefit can be quite profound.
To begin, you might experiment on a small and well-defined task. Start with a prototype to prove the concept before committing to a larger investment. It will assess the likelihood of success and engage your internal stakeholders, comforted by the results.
Starting small is also a good way to learn about two potential risks of AI: transparency and safety. First, outcomes of machine learning are not always transparent for the user. Second, deep learning models are highly nonlinear and can generate surprising results with minimal changes. An image classifier can be completely fooled with a well-chosen, very small overlay, invisible to the human eye.
To remedy these potential risks, your AI, if embedded in a decision process, will most likely need a high level of supervision at first: a human behind the machine validating its results.
Having some history and monitoring performance in the supervised phase will build the confidence needed to eventually remove constant supervision completely.