AI and Banking: Competing with total trust and zero bias

Share this post:

A recent article in The Times highlighted that only half of UK companies have a clear AI strategy in place. There is huge potential for companies across industries to leverage the power of AI to improve their business. The banking sector is no different. But for all its benefits, what makes AI ethical? How do you address these issues around it? As decision-making using AI becomes more accepted, how can banks ensure that AI can avoid bias and discrimination? And how can the technology coexist, both with the Human factors at the bank, as well as fit into the brand and ethos.

Recently I had the pleasure of attending NextGen Banking event hosted by FinExtra in London where I participated in a panel discussion, as well as getting to talk with clients about IBM’s offerings in the Financial Services Sector. From front office to back, it was clear that banks from across the industry are really excited at the prospect of implementing AI and harnessing the benefits it can provide.

The panel discussion was centred around ethical issues and AI – how can we use AI in the industry with complete and total trust that it will eliminate bias in the processes it delivers. Alongside me on the panel were: Ekene Uzoma, VP Digital Product Development, State Street; Monica Monaco, Founder, TrustEU Affairs and Terry Cordeiro, Head of Product Management – Applied Science and Intelligent Products, Lloyds Bank. We answered a variety of questions from the audience. One of the questions was about the gender identity of virtual assistants. As one of the most common front office applications of AI, they allow the bank to service clients’ needs more easily, helping to assign and utilise human agents to the cases that need the most help and interaction. The question raised highlighted that a significant number of virtual assistant services deployed at companies across industries were predominantly personalised based on a female persona, implying this was an unconscious bias. This was even before you get into the bias inherent in models, training data and team assumptions.

See my short interview with Finextra for a glimpse of what was discussed below.

What is encouraging however, is how seriously the industry is taking these questions and really setting themselves up for answering them properly and robustly.

I am delighted to announce that to coincide with IBM’s presence at NextGen 2019, we are releasing a white paper, entitled “Controlling AI: How to regulate and safely exploit the proliferation of AI” which you can find here.

With this being such a nascent field, I would love to debate this further and hear other’s points of view – so please feel free to reach out to me on Twitter to carry on the discussion in this hugely interesting new area.