‘People who bought this swap also bought Apple’

This opinion piece accompanies our series of four feature-length articles on front-office digital change. The first two features are available here and here. The third will be published on July 12.

There’s a video on YouTube introducing a new, Natixis-built bot. Designed as an add-in for bank-owned messaging platform Symphony, it illustrates how the industry plans to turn chat into an all-singing, all-dancing workflow tool.

Equally interesting – though for different reasons – is the queue of autoplay videos YouTube stacks up on the right-hand side of the screen to keep me watching.

I probably shouldn’t tell you this, because the videos have been selected on the basis of my personal viewing habits but, after the Natixis presentation, we leap quite quickly to cognitive biases, the “10 worst British military aircraft” and a short film in which a small girl gives money to a street performer and “gets the best surprise in return”.

You’ll have to take my word for it when I say I’m not particularly interested in any of that. At least, not right now.

I bring it up, because personalisation is another part of the industry’s plans for a tech-led revolution in its approach to sales. The idea is that self-educating algorithms will listen in to phone calls and chats between banks and their clients, pick out and understand the clients’ needs, and then run off to fetch something useful – market data, say, or relevant research, a trade idea, or a product the client might want to buy.

The call might even be happening because, based on recent market moves – and the bank’s knowledge of the client’s positions – the algorithm suspects now is the moment to pitch the client a particular trade and prods the salesperson to pick up the phone.

Banks believe they can make this work. But that may be because they need to make it work.

With large swaths of the buy side facing cost pressures of their own, and more business being done electronically, dealers anticipate a shrinking wallet and tighter margins. If they want to keep these customers, automation is the way to do it – and a computerised assistant that can perfectly anticipate client needs would solve everything.

If a bank keeps calling you about a product you don’t want, then you stop taking their calls

Electronic trading head

But the problems that lead YouTube to think I care about cognitive biases and military aircraft also afflict dealers: they don’t have complete information about what their clients want. They only see the trades and enquiries that come to them; they don’t know anything about the trades that go elsewhere. And they don’t always have the context: is that swap a hedge of an existing position or an investment view? Is the client free to take profit or not?

Amid all the excitement about this idea, one electronic trading head sounds a note of caution:  “When you’re on Netflix or Amazon, and they recommend something you think is completely useless, it doesn’t matter – you’re still going back. But if a bank keeps calling you about a product you don’t want, then you stop taking their calls. So although these techniques have been around for a while, they’ve never been required to generate the level of precision we would need in order to feel comfortable using them,” he says.

He’s not writing the idea off, but he envisages “multiple years of testing and refinement” before the tools are in widespread use.

That sounds realistic. Although maybe I’m a victim of confirmation bias. I’ll just look it up on YouTube.