Artificial intelligence and natural reluctance
I love Sibos. Have I mentioned how much I love Sibos?
Seriously. At the risk of sounding exactly like the rapidly ageing nerd that I am, I love Sibos because it is a grown up conference. Despite the swag and daytime drinking, Sibos is different from all other industry events for two simple reasons: firstly, the content is mature, complex and often dense. Complicated or not, if we as an industry need to know it and think about it, we will talk about it. Be it Target-2 Securities (T2S), capital adequacy and fiduciary pressures or, indeed, the future of money. Secondly, the attendees are mature. They are accomplished professionals with their organisation’s clear mandate and the power to make decisions. It makes for an altogether different animal than your average conference populated by marketing bods, interns and people with “ecosystem” in their job titles.
Every year I get to achieve something and learn something at Sibos, which is more than I can say about most weeks and any other conference.
And this year was no exception. Among other weird and wonderful things, I got to chair an AI panel with three incredibly brilliant people full of stories, insight and the experience of things going wrong before they go right. If ever in doubt, that is the surest way of testing whether someone really has the battle scars they claim they have. Their stories go wrong before they go right.
As we were prepping for the panel, it was clear from the second we started that my panelists had way more to share than the time allocated to our session would allow. And before you go “d’uh, of course” let me state this is not always the case. It is not uncommon to have a so called “expert” on stage whose entire body of knowledge on the topic they are pontificating on is shared during their 15-minute slot in the spotlight.
Not these folks.
They had stories. They had insight. They had perspective.
Don’t spook them
The first thing I told my panelists by way of prep was don’t spook the audience.
Don’t lose them.
Don’t go too technical, too futuristic. Give them something they can sink their teeth into. Give them something relatable and accessible.
And they did. They talked about AI helping us sift through masses of information and detecting patterns that were always there but the human mind couldn’t see them through the vastness of detail: they talked about realising that we, as an industry, are many times more likely to flag a transaction as suspicious if the client is called Mohammed than if he is called Maurice. AI helped us see the bias, humans helped us address it.
They talked about how AI can scan through x rays and prioritise the ones that don’t look healthy on the basis of precedent and pattern recognition, the ones that need urgent care while sending other patients home with the reassurance that all looks fine and the doctor will be in touch in due course. So nobody falls through the cracks.
Relatable stories. Real examples.
They talked about the dangers of algorithmic bias but also the ability of AI to help us detect the bias we already have. Human and inevitable. They talked about AI being a handmaiden rather than a tyrant, freeing humans from tedium and enabling them to deal with nuance, exceptions, value-additive work in banking or aviation, healthcare or administration.
And they talked about the very human challenge of coming to terms with what has only been recently possible. Digesting. Believing. Trusting.
And they talked about the fact that we have a long long way ahead of us in this space. To grow and learn and develop capabilities.
The second thing I told my panelists by way of prep was spook the audience.
Scare the living daylights out of them.
We are bankers. You don’t want to tell us “there is a long way to go here” because all we will hear is “this is the next guy’s problem, you have enough burning platforms on your hands so you can allow yourself the relief of not worrying about everything”. If the technology is years away from being mature, the banker is light years away from being engaged. Don’t lull them into a sense of security because it would be false. This technology is very much here. And although there is a long way ahead of it towards maturity, there is a long way behind it too and it’s more than able to kick ass today, thank you very much.
It is here. Don’t sit back.
You need to know because your competitors are increasingly in the know. And you should be spooked. Because this is not easy.
This is not a little play on the sides of your desk. This is not a painless little pilot.
An AI experiment needs a ton-load of data, serious amounts of thought around the type of data organisation you will become. It needs new talent (data scientists are hard to come by and they travel in packs). It needs a new kind of management mindset. It needs a new kind of problem solving mindset. And you need to mean to go on, if you start on this path. Otherwise the effort is too high, the complexity too taxing.
And before you are tempted to go “d’uh” for the second time in the space of one blog, let’s be honest with each other. How many banking experiments are designed for success? Banking innovation is either experimentation expected to never see the light of day (even when it succeeds against its stated criteria) or watered down to irrecognisability in the name of risk mitigation. A new thing twisted to look so much like an old thing that you find yourself wondering why we bother.
Not so with this technology.
It needs scale and depth and conviction.
Your smallest experiment needs to be big enough for the data to have critical mass to yield results.
You need to harness resources. You need to focus the organisation.
It’s hard work. It’s a lot of work. It does not bring immediate results. You have to get your approvals and blessings. You have to find the people. You have to train the people. You have to select, release and cleanse the data. You have to actually programmatically design. Then you have to train your algorithms. The machine doesn’t know before it knows. It takes time to learn. Time, data and humans doing a lot of modelling.
And when you succeed, and you will, you need to have the willingness to become the sort of data-driven organisation that will go back for another round of this as soon as you are done. Because what is the point unless you are willing to become a smart organisation, a data-driven organisation, an organisation whose heart beats to a digital pulse?
And all this has to be done at the same time as funding your APIs, revamping your core, designing new propositions and experimenting with whatever new thing comes down the pipe and dealing with the regulators’ latest missive, not to mention running the business and staying ahead of the competition in all your core markets.
The prize is immense: building a future-proof organisation with a right to not just survival but also success.
Nobody said it would be easy.
As always, up close, opportunity looks like hard work.
My panel did their job beautifully: inspire, incite, terrify.
The future does not come easy and success does not come free. Time to get to work.
By Leda Glyptis
Leda Glyptis is FinTech Futures’ resident thought provocateur – she leads, writes on, lives and breathes transformation and digital disruption as CEO of 11:FS Foundry.
She is a recovering banker, lapsed academic and long-term resident of the banking ecosystem.
All opinions are her own. You can’t have them – but you are welcome to debate and comment!