FinTechAutomationGresham Tech CTO: AI is great at stating the obvious

Gresham Tech CTO: AI is great at stating the obvious

CTO of Gresham Technologies, Neil Vernon discusses the future of technology in the post-lockdown world

The Global Treasurer spoke to Vernon to understand the challenges that firms face as they move out of lockdown, the importance of humans in the tech revolution and where artificial intelligence (AI) and robotic process automation (RPA) will take us over the coming years.

As we come out of lockdown, what do you see as the biggest challenges that firms will face?

If we look back at when the markets crashed, the markets crashed around about March 12 or 13. The banks we were dealing with saw huge volume increases, as in 30 percent more volume or 40 percent more volume. But actually, that was volume going through systems that were already on the edge of not being able to cope. So, when you added what is certainly not that much, big things started to go wrong.

What happens when things go wrong is that your controls infrastructure starts to spot that those things have gone wrong. For one of our biggest customers, and one of their main controls in the back office, typically every day they will get in and find 6,000 transactions that had some kind of problem. A transaction that had failed to get from middle office to back office, or a transaction that had failed to get enriched because a lookup failed, or a transaction that was inconsistent across the infrastructure.

So, on a typical day there were 6,000 problems neatly categorised, some actions to be taken on each of them or a group of them. But on March 13, that organisation had 600,000 problems. The volume of transactions had gone up by 30 or 40 percent, but the volume of exceptions and problems had gone up by 10,000 percent.

When I was talking about this, the conversation was ‘can AI help you there?’ And my sense at that time was the bank knew that their systems were failing. They knew that the control tomorrow would show lots and lots of problems. They already knew, because of their knowledge of the infrastructure, what the right solutions will be for the majority of those problems. In fact, for at least 500,000 out of 600,000, all they had to do was wait. They knew their systems had got into this massive backlog and if they waited 12 hours, then a lot of the problems would just go away.

But people were advocating AI in these situations and not actually leveraging the human knowledge that already exists. And I think we do ourselves a disservice if we think that an AI will replace a human.

You could have put an AI on those 600,000 problems and it may have found some things that were interesting. But it was already known, the human knowledge already knew what the problem was and already figured out a resolution plan before the problem was highlighted by the control.

So, to your question, for the future I think spiky days will never go away. If anything, we can imagine spikier days.

We are always going to have to deal with exceptional days, where exceptional in our world means there’s lots and lots of problems, far more than you had yesterday. You’ve got to be able to use your knowledge of your infrastructure, knowledge of your process, knowledge of your industry, and knowledge of your business to be able to foresee what the effect of those problems would be and then have a plan to deal with them.

And that’s what we’ve seen with our customers, those that had the most knowledge about their processes dealt with the problem really well.

How are you seeing your customers adapt to this ‘new world’?

I think there’s an element of getting real and I think that’s really important. The glamour projects are being cut and there’s a real focus on what I would say are the more important things – that focus on what solutions there are for today and delivering those today.

Don’t get excited by technologies that are as yet, unproven when there are real solutions to actual, real problems now. There is always a focus on ROI and making sure that you’re making the right kind of investments. But I think the riskier investments are being seen for what they are at the moment, and it’s low risk strategies that are important.

I was going to use the phrase low risk and low return, but I actually don’t think that’s the case. There’s still a lot of low hanging fruit within a low risk strategy that can still give you a high return.

Do you think there’s almost too much of a focus on these extremely high-tech things that aren’t yet ready? And we don’t trust the humans enough?

In a nutshell, yes. I think they’re not yet ready. They are an unproven and you can waste a significant amount of project lifecycle on tech like that.

You don’t need to put AI in, you need to apply your brain and you’ll know when you have problems.

I see that time and time again that the AI seems to be really good at telling you the bleeding obvious. And you can waste a lot of project cycles and lots of money to learn what a knowledgeable person in the organisation already knows.

If we flip that around then, what can AI and RPA do to help companies out of the crisis?

Well, I am a big fan of RPA and I’ll caveat that in the moment. But I think this idea that there are these repetitive actions that humans go through, where we categorise assessment transactions into a problem space. We make sure they all conform to that problem space and then we apply a consistent action to that set. That can get automated time and time and time again through RPA.

If the RPA is embedded in a product, it can give real value and create greater job satisfaction because there’s nothing more boring than doing the same thing over and over and over again. So, I’m a big fan of RPA. I think organizations can invest in RPA and get value from it.

We [Gresham] are on a journey with that ourselves in that we’ve seen the value of embedding RPA techniques into our products. But there are two downsides to RPA I see at the moment. The first is that if you use an external RPA tool, then those robots tend to break whenever there’s a vendor upgrade because the vendors don’t have knowledge of how the robots are driving the screens. Your infrastructure can become a bit frozen in time.

If you just naively adopt RPA it gets frozen in time because everybody’s scared to upgrade, because whenever they upgrade the robots break, so upgrades become more and more expensive, and less and less frequent. Of course, the irony here is, the less frequent they are, the more likely they are to be problematic when you do upgrade.

Then secondly, my other downside on RPA is there is a temptation to use RPA for things that it shouldn’t really be used for. In particular, I’m thinking about integration on the cheap. If you want to connect system A to system B and be able to do that volume at pace there is no proper way of doing that other than through proper integration.

So, I do think there’s a big caveat on RPA but there’s a lot of value to be had with it too. Just carefully apply it and don’t over apply it.

Do you see the human becoming more important as the technology gets better?

I don’t see any inkling of a clue that human insight will not always be valuable. No matter what we do with the technology, that insight will always be better than the AI. I don’t see any evidence points at all for that. And I say this is as a vendor that’s got AI techniques in our products.

There’s an Emperor’s New Clothes element to it. You can wow people with it, but somebody that really understands the business will say ‘so what?’


Whitepapers & Resources

2021 Transaction Banking Services Survey

Banking 2021 Transaction Banking Services Survey

CGI Transaction Banking Survey 2020

CGI Transaction Banking Survey 2020

TIS Sanction Screening Survey Report

Payments TIS Sanction Screening Survey Report

Enhancing your strategic position: Digitalization in Treasury

Payments Enhancing your strategic position: Digitalization in Treasury

Netting: An Immersive Guide to Global Reconciliation

Netting: An Immersive Guide to Global Reconciliation