Meet your new algorithmic overlords 13 Oct 2016
Artifical intelligence does not have such a great reputation in our culture. HAL 9000, Skynet, that big glowing cube thing in I, Robot… Let’s just say we humans have something of an inferiority complex about machines doing the thinking.
But that’s not going to stop those machines coming for your finance function. In the words of tech blogger Vik Singh, “AI is hot, I mean really hot.” He cites the $1.5bn that VCs invested in AI businesses in the first half of 2016, compared to $2.4bn for the whole of 2015 and just $282m five years ago. That's a lot of bets on a technology that will need to make a big difference in all walks of life.
Let’s pause for some definitions. According to engineering lore, a computer must pass the Turing Test if it’s to be considered “intelligent” – that is, it should be indistinguishable from a human in conversation. (Alan Turing actually calling this the “Imitation Game”.) Today, the preferred catch-all term is "cognitive technologies", which IBM defines as “using natural language processing and machine learning to enable people and machines to interact more naturally to extend and magnify human expertise and cognition.”
Machine learning means software that gets better at doing what it does without being told how by humans. Natural language processing means they understand us, our actions, even our feelings, better. So the algorithms that make 75% of the trades on the world’s capital markets are, in a sense, AI because they learn and adapt to circumstances without intervention. (Although just because they’re “intelligent” doesn’t make them smart or safe – witness the recent algo-driven flash crash in sterling. Not that it needed much help, mind you…)
Understanding complex systems and, especially, predictive modelling are thought to be the killer applications for AI. Hence their application to areas like HR (people are complex; companies need to know how many they’ll need) – and finance.
There are lots of articles about how many of the transactional and even analytical functions currently done by people in finance are going to get done better and faster by cognitive technologies. And it’s true that smart FDs and up-and-comers need to get on the case. Google needed a supercomputer to beat a human at Go in March – but similar predictive and even quantum-based technologies (that stuff is really mind-blowing) will be part and parcel of bog-standard ERP within the span of most of your careers.
But Singh’s blog on the subject is a must-read because he puts a crucial context around AI in the workplace. He called it AI’s “last mile problem” (a reference to the issue deregulated telecoms companies have with the last mile of cables into people’s homes.). He asks:
“How do you get regular business users to depend on your predictions, even though they won’t understand all of the science that went into calculating them? You want them to trust the predictions, to understand how to best leverage them to drive value, and to change their workflows to depend on them.”
I did some work for KPMG recently where they mapped a whole host of finance function roles and asked what would be left for humans once the AIs get to work. By 2020, they reckoned, even very human activities like “coaching” and “managing risk” will be semi-automated. “Developing commercial insight from financial analysis” has, by then, hit the “fully automated” column. By 2025, only “coaching” and “managing projects” have escaped full automation.
According to the World Economic Forum, AI will really start to kick in from 2018. That creates a window of opportunity for organisations to consider who they want, in which positions, doing what roles as this revolution plays out. And for the finance function, that means focusing on complex problem-solving, critical thinking and creativity. (Although, cheerily, the WEF report says the big hits will be manufacturing and admin: "Other sizeable job families, such as Business and Financial Operations, Sales and Related and Construction and Extraction have a largely flat global employment outlook over the 2015-2020 period." Gee, thanks.)
As for Singh’s last mile problem – well, let’s couch it as a boardroom problem for the finance exec. “Should we proceed with my brilliant plan to acquire our rival in Germany,” asks the chairman. As FD, do you have the courage to reply simply, “Computer says ‘no’.”? When that happens, we’ll know AI has really arrived. And you’d better have something interesting to add to the algorithm’s verdict when it does…