i-robot

A spectre stalks the land of banking employment – the spectre of automation. Of course, it isn’t a spectre that is confining its scary stalking to banking. Try speaking to a London black cab driver about the threat to ‘The Knowledge’ and thus to his livelihood from ever-smarter Satnavs and from apps like Uber if you would care for a high-pitched, high-volume exploration of this fact. But bankers (and other finance sector workers) are particularly vulnerable.

First, their jobs are primarily concerned with processing abstract, ‘virtual’ information. This marks them out from, let’s say, hairdressers who need to operate in physical, base reality and puts their activities squarely into the realm in which computers can compete with them directly. Second, bankers tend to be rather expensive. Not everyone earns the millions so entrenched in popular imagination (in fact, only a tiny fraction do) but still, collectively, they are a huge expense for any financial firm. HSBC, for instance, spent almost 20 billion pounds on employee compensation and benefits in 2015 – significantly more than its post-tax profit of 15 billion. Actually, this is probably an underestimate of the true cost of paying wages since HSBC also incurred 20 billion of ‘other expenses’ a chunk of which is payment to people outside the firm (e.g. consultants, legal experts et al). [1] Other banks have very similar ratios.

Thus, over the years banks have made strenuous efforts to automate jobs out of existence. Much of this effort on the retail side of banking has been spent on the ‘back office’ activities of transaction processing – routine, often repetitive work that can be fairly readily computerised. The rise of Internet banking is another symptom of this, a development that also has the advantage (seen from the point of view of bank shareholders and senior managers) of reducing the need for so many front office, client-facing staff.

The really expensive people in any (universal) bank, though, are its investment bankers. These are the people on whom, if not millions of pounds, then at least hundreds of thousands routinely get spent each year on salaries and bonuses. Can their jobs be automated? The plain answer, in many cases, is ‘yes’. We have already seen swathes of jobs for traders of cash equities or spot FX eradicated by the use of computers. These simple, homogenous products lend themselves to automation.

In the future, as computers get progressively more powerful per unit cost, the temptation for banks’ managements will be to extend this approach to more complex products. The critical trade-off in the mind of any manager making such a decision is a simple one: how feasible is automation (a function of the complexity, fragmentation and ‘wrinkle-density’ of any market)? And is it worth it (the bigger the market, and more expensive to staff, the better since more money can be saved)? With this in mind, over time I would expect to see the gradual automation of market making in products like interest rate swaps, credit default swaps and FX forwards since the balance in favour of automation is, or will become, compelling. All three markets are gigantic, relatively straightforward (as compared to complex derivative markets) and employ expensive traders. Even if banks are too distracted or resource constrained in terms of available technical folk or of balance sheet, my guess is that hedge funds will step in. Indeed, a spin-off of American fund giant Citadel is already flexing its muscles in interest rate land. [2]

So much for traders: what about customer-facing investment bankers? At the high end (advisory work on mergers and the like) jobs are safe. The senior roles are complex, involve a great deal of human contact and are essentially non-automatable. (Though whether that applies to the army of juniors who provide support and assemble hundredweight piles of vital, but yawn-inducing documentation is another matter.)

Then there are the salespeople employed to speak to customers who wish to transact in the bond, equity, credit or FX markets. In this number, the folk who just offer purely transactional, execution services (pick up the phone, receive an order, get it done) already are, or surely will be, toast. Computers can do the job better. The remaining people must offer something more – advice, ‘colour’, expertise. How easy is that to replicate? Leaving aside the inconvenient fact that regulators are tightening up on exactly what information salespeople can give out (since one man’s ‘market colour’ is another’s ‘leak of sensitive order flow information’) I’m not sure how difficult these skills are to mimic with technology.

Early in my career I came across an FX salesman with a rather brilliant approach to client advice. The client would be long 100 million USD against the German mark, say, and would ask my colleague, ‘what should I do?’   Inevitably the answer was, ‘sell half’. Then, if the USD rallied, the salesman would point to the gains on the remaining half of the position. If the USD sold off, on the other hand, he would major on the losses he had saved the client by advising the sale. You don’t need to be a PhD in computer science to figure out how that rule of thumb might be programmed.

Clearly, by bringing this man up I am being flippant. But the truth is that the range of potential advice that can be given to customers in all but the most dramatic circumstances is pretty limited. Couldn’t computers, linked to the latest research, able to data-mine market patterns in seconds and with a live feed of order flow and the details of a customer’s positions be able to do a job at least equal to that of a human?

If they can, then I suspect that, in time, they will. Sad to say, the next wave of the unemployed from banks’ trading floors may well come from ranks of their most charming inhabitants – the sales force.

 

[1] HSBC 2015 Annual report, http://www.annualreports.com/Company/hsbc-holdings-plc

 

[2] Citadel pushes into European swap markets, Financial Times, 10 September 2015, http://www.ft.com/cms/s/0/319edd98-57cb-11e5-a28b-50226830d644.html#axzz4DQv5DjGc