The age of APIs
Having attended several years of Money20/20 in Las Vegas, I have seen many topics coming in and out of fashion. In 2016, I recall that APIs, open banking, and regulatory-driven data initiatives were hot new things. Yes, blockchain was there too, but by then crypto was already trending downwards for sensible attendees (in direct counter with eyebrows). Peak ledger arguably had already happened the year before.
A year later, in 2017, the rise of artificial intelligence (AI) was noticeable. Companies with AI in their title, or products enhanced by machine learning were the conference stars. APIs hadn’t disappeared as a theme, but more people were talking about how best to implement open APIs rather than asking “what the hell is open banking?” Among all the usual glitz around new Point of Sale (PoS) devices, designed to look like Apple stores, there seemed to be an underlying transition from API to AI. Essentially, things were moving from functional connectivity to smart agency.
The risk of AI fatigue
But as we hurtle toward 2020 (the year, not the conference), AI may already be suffering from hyperbolic overuse. It did not take too much curiosity to expose that some conference exhibits claiming to be powered by AI seemed to be doing no more than accessing “what-if” spreadsheet macros. I’m not anti-spreadsheet – debugging Lotus 1-2-3 macros is surely a key skill I’ll be able to fall back on one day – but that’s not intelligence. Not as we know it, Jim. Also, some industry participants haven’t helped the cause by claiming use of AI within their “smart” chatbots, voice recognition processors, email servers and app engines, only to be exposed as largely relying on human intelligence (i.e. real people) to manually manage the work!
How do we rise against this machine learning malaise? We need more AI and machine learning capabilities in solutions that can be implemented in a step-wise manner. Small, tactical projects with a return on investment will help financial institutions make the case to scale human resources in line with their systems, for example in fraud detection and prevention.
What I’m seeing in 2020 are huge advances in processing power (exemplified by quantum computing), the rise of autonomous devices within the Internet of Things, and access to ‘Big Data,’ which is certainly fueling capability advances in computing intelligence. New AI techniques are based on finding and extracting patterns within seemingly unlabeled and unstructured data, finding relationships, meanings and significance for the benefit of the AI user. Financial services systems are certainly evolving from complicated rules-based machines toward the realm of “intelligence.”
But it is worth pausing to wonder about a typical AI user. Is a self-driving vehicle a typical user? Or an advertising agency’s revenue management engine? Or perhaps a bank’s compliance machine? Surely the latter would be more efficient for an AI entity rather than a human to navigate the policies, regulations and interpretations needed to avoid modern regulatory pitfalls!
Without being too precious about purely academic definitions of AI, it seems sensible to embrace Asimov’s laws of robotics, which, in summary, dictate that this new AI must always be programmed to protect humans. By extension, AI needs to be driven to work for the best interest of individual citizens.
The consequences of this approach can be visualized by imagining your own personal avatar, operating as your agent in the digital world, where you have specified the controls and limits of what you would like your agent/advocate to do for you, and the boundaries beyond which you want to retain physical, hands-on control.
So, the “I” in AI – what does it stand for? “Intelligence,” I hear you cry. But perhaps it should symbolize a more important concept? Perhaps it would be better if ‘I’ were to stand as it is – in its abbreviated form, for I. As in, “me.”
Me, myself and identity
And what makes I or me, me? We need to come to grips with identity. Not in the traditional sense, but more in terms of validated permissions and entitlements that allow an individual to remove any doubt of that individual’s rights to access a particular service, or to initiate specific transactions.
This requirement for ID-based, strong customer authentication linked to me/I is already a reality (if in doubt, see the Regulatory Technical Standards of the revised EU Payments Services Directive). The protocols of PSD2 may appear clunky for those looking for friction-free payments – it may feel wrong to challenge a user with a “you are about to pay $X to recipient Y; please enter your credentials one more time to confirm this is OK” requirement. But, as our lives become increasingly digital, these new ceremonies of ID verification (hopefully simplified via biometric techniques) will become an essential part of life.
Although ID features heavily in current conferences and exhibitions, we are still in the early days of embracing I – so we still have a way to go before we hit widescale adoption.
Much work needs to be done to merge new authentication requirements into personal lifestyle management systems. The current complexities faced by organizations responding to policies and regulations are tricky to resolve, but a shift toward focusing on how individual’s needs will be catered for will help. Remember, this is all about me. I is me – it’s my life and my data. I expect a lot more offerings geared specifically for me, not a continuous bluster of mass marketing generated for the benefit of a big data hoarding company.
We know what B2B and B2C are all about. But now we need to move on. Let’s think about “Me2B.” That’s proper customer-centric thinking. With me, my identity and my preferences are well-presented and accessible only to businesses that can do something useful for me.