“If you have a complete system which automates risk management and platform decisions from ground up, not knowing their role, then you’re choosing the wrong provider. AI is not here yet, so if you are choosing someone who has built the all encompassing AI solution from front end to back end, you might want to look at a firm that’s been around longer””
Within large institutions recently, the adoption of artificial intelligence (AI) solutions to replace entire departments and subsequently cause the redundancy of staff has sent a wave of modernity through the financial sector that has been met with some degree of trepidation.
Recently, FinanceFeeds reported that Russian financial institution Sberbank replaced 3,000 staff with an AI solution that operates its legal department.
This may well be a direction within the large banks, however the conundrum that still surrounds automation and machine learning in the OTC FX industry is still something upon which the jury is out.
Today, here at the iFX EXPO International in Limassol Cyprus, produced by ConversionPros, which is host to over 3,500 pre-registered attendees and 500 further delegates which arrived during the course of the event, a debate which revealed opposing opinions ensued during the early hours of the afternoon, the upstart pro-full automation battling it out with the experienced developers of key components that are in use on a widespread scale globally within several hundred electronic trading firms.
Introduced by Finance Magnates Editor in Chief Jonathan Fine, discussions commenced. Mr. Fine explained “Three years ago, social trading was a buzzword within this sector, many firms using it at several opportunities. Today, the buzzwords are AI and automation, therefore we would like to look at how AI can drive the modernization of the FX business, and of course very importantly, reduce costs.”
Moderated by Steven Hatzakis, an interactive discussion held by professionals who know the industry and are certainly qualified to interject where necessary polarized opinions when it comes to the role of intelligent computers in this sector.
Mr Hatzakis said “AI has been gaining a lot of attention, and is in many generic technology orientated news feeds. I would like to start this discussion by working out what this means and who is using it, our distinguished panelists are going to elaborate on this. Is AI a buzzword, and how can we distinguish between automation and AI, as without automation we wouldn’t have AI. Therefore. where is the line drawn between automation and actual AI?”
Jonathan Frankenstein, Director at OSYSTEMS began to answer this question.”There is a distinction between them. We use a lot of marketing materials such as automated emails. When I look at AI, I look at more deep data, along with the algos, learning and training. This is something that can help with all of the call centers within brokerages.
Mr. Hatzakis then asked “Where do you see AI on teh CRM side? That is an important process within brokerages”, to which Mr Frankenstein said “Yes, With AI, we will be able to give the sales person the higher quality of leads when it comes to conversion and cut some of the sales force. It is all about optimization.”
At that point, Gil Eyal, Co Founder and President of Tradency interjected, saying “Coming from theoretical computer science angle, the differentiation between AI and automation is clear. When machines are built there is a set of functions that machines are doing better than humans, such as simple calculations, memory, and repetitive processes.”
“The handling of these tasks is something to automate. Automation can now take the role of cognative human tasks, where machines can talk with each other, imagine things, and perform lots of tasks in which machines are trying to imitate and do as much as possible” he said.
“Our product was based on this methodology” said Mr Eyal. “At first, the idea was to use servers that operate 24/7 and optimize in a certain way. We wanted to give retail traders the same ability to trade, even when sleeping, eating or working that they would have if they were trading manually, therefore we designed it so that the servers are in the market monitoring strategies and then execute on behalf of the user.”
Today AI has evolved into decision making, defining risk, leverage…
Mr Hatzakis then addressed Andrew Ralich, CEO of oneZero Financial Systems, asking “Andrew, where is AI most effective?” – Mr Ralich then answered “I wouldn’t be in this industry if it wasn’t for automatiomn. This is what we develop. Automating what needs to be automated is essential. Where automation has grown is similar to what Gil mentioned – telling a machine what to do. For example, to decide whether a specific execution is A book or B Book.”
“I draw the line between advanced automation getting into AI in which I would say that hedging risk is simple automation, whereas identifying a scalper is a more advanced form of automation. The innovation that we are seeing today within brokers looking for within systems is that for example, a client just made that much money for a broker, so the broker can then adjust the risk management strategy as a result of a decision that was made form within the technology.” – Andrew Ralich, CEO, oneZero Financial Systems
Mr Hatzakis then asked “Have you seen AI show up in that yet? Has it reached the point where it is machine learning? “It is a difficult question” said Mr Ralich.
“At a theoretical level, AI implies machine learning. I am not sure if within this industry you don’t have many situations where the risk management system is identifying a new pattern of trading where a trader can be moved from book to book. There is a lot of progress to be made with regard to that type of reporting from human action” he said.
Mr Hatzakis then posed a point with regard to how to solve this challenge.”Oded, “What are some of the straegies you are seeing become automated?” Oded Shefer, CEO of cPattern then explained “I wanted to give an example from the world fo fighting terrorism on the distinction between AI and automation, then I can explain what brokers are doing.”
“In airports, 20 years ago, if you wanted to have security, you’d have to have police officers on the ground looking for suspicious movements which was human orientated work, however now they put cameras everywhere and have one person monitoring the cameras. Removing the humans with ‘cameras’ that have no wisdom at all, replacing security officers – this is automation, whereas AI is monitoring the crowds, looking for faces that are recognizable as suspicious and making the right action accordingly” – Oded Shefer, CEO, cPattern
“What I see is that there is the task of a particular employee within the back office of a brokerage. Automation is where you take specific work processes that take over, and this is gradually becoming automated, and now it has sophisticated actions. I consider this to be advanced automation, where there is wisdom in the system, wherever it is identifying tasks that a human cannot do in a simple way, this is AI.”
Mr. Hatzakis then said “Are you mirroring what Andrew said in that we arent there yet?” Mr. Shefer then explained “It is a process. Automation and AI is a process and evolution that takes time.”
Gilad Gat, Co Founder of TipRanks said “We operate a lot in the equities and stocks area, which is an important sector to look at with regard to the use of AI if you see things being moved automatically. We are seeing portfolios being moved automatcially on the wisdom of the crowds. Some of this information harder for us to analyze without using machine learning.”
“We analyze every article that is in the financial publications. We then wanted to expand on it, for example with regard to which stocks should be bought or sold. Conventional means of automation were not enough and what people write is so unstructured and is hard to quantify and model, so we moved to machine learning to understand when someone is writing about a certain stock, is it going up or going down and this way we can move data to customers” – Gilad Gat, Co-Founder, TipRanks
Mr Hatzakis asked “Can you then use this to research keywords?” Mr. Gat replied “We look at paragraphs and articles and understand from that what is the quality of the offer.”
Mr Hatzakis interjected and asked “Does it improve based on experience when measuring its performance?” to which Mr Gat responded “To improve the machine learning we have to constantly give it feedback. We analyze the articles from financial bloggers, so it gives recommendations with accuracy, for example ‘this is a bullish recommendation but I’m 75% sure, or 100% sure.’ We then have someone manually go over them to ensure that this has been done correctly.”
“Rather like Facebook going over its content to check for fake news?” asked Mr Hatzakis. Mr Gat replied “I think that’s a good example yes.”
What is lacking with regard to firms needing to get on the AI bandwagon?
Mr Frankenstein then said “Take KYC. Now we have a 3rd party provider which makes compliance checks on the spot so that it can be established whether documents are valid or not. Brokers get a response about it immediately, which saves time, requires less people and takes the error aspect out of it. When it comes to AI, on the call center side, the prediction the marketing funnels – the machine learns it, then it gives the sales person leads from top to bottom. This is something that is learning according to accurate data.”
Consensus: AI Today is not there yet, it cannot live without automation and the sample size of data isnt there. Once proactive decisions appear then we will be able to assess data and AI will be much more ergonomically sound.
Mr Hatzakis then asked Mr Ralich “Andrew, where can automation have its biggest impact, on your side of the industry in integration, automated risk management, is there room for AI there yet?”
Mr Ralich said “It hasn’t got to that stage yet, and even the advanced automation hasn’t yet. I am certainly here to tell you that this is not what we are working on at oneZero. The prices of building AI into the back end is not the role of the order management companies, execution specialists and integration firms.”
“It is worth looking at the Tesla car. This is a good example of top level automation but the sustem that is analyzing is not part of the braking system or the engine, it is a separate component that is interacting with the other key components” – Andrew Ralich, CEO, oneZero Financial Systems
“The investments that we are making are in API, and toolsets that allow us to tell brokers to achieve their end progressively. We don’t want to do be doing this analysis within the core of the platform” he said.
Mr Hatzakis responded by asking Are proprietary traders using genetic program and neural networks to trade and manage risk? Is that a secret sauce? Can this be a game changer for the industry, and with MiFID 2 coming up will AI have a role driving price discovery?
Mr Ralich responded “The responsibility of technology vendors is not coming up with the genetic algo that exists in real time. Mostly, firms ask how can we learn from the past? We invest a lot in the dataset, the parallel processing that will facilitate this real time analysis will be groomed around a data set that we are developing today. Even in the higher end of the non bank sector such as Citadel Securities or XTX, there are genetic algos, however most is being done on historical data and then being hard coded in. I don’t think anyone runs a real time algo now.”
Where is it lacking on the AI side?
Mr Shefer said “I think you need to look at the traders and their needs. When they invest money they need to have trust with the broker. If I was a trader, which I am not, but if I was and I invested $5000 with a broker I would want a human speaking to me rather than an automated message.” Mr Hatzakis said “You are not a Millennial!” to which Mr Shefer convivially responded “No, I’m an 80s person!”
“There is a big spectrum of negotiation between client and borker that will be hard to substitute with AI, but 60 to 80% of interactions between broker and client could work” said Mr Shefer.
“By this, I mean not improvement but coverage between trader and broker, including lots of contact points between client and broker. These could be chat, log in, log out, depositing, withdrawals, lots of touch points between broker and trader, this is a huge amount of work and data. I think we can see gradually brokers covering more and more actions that could be replaced by the computer. Of course you don’t need a human to be there at every point of time but there will always be 20% of the interaction that should be human” – Oded Shefer, CEO, cPattern
Mr Frankenstein then explained that he considers customer engagement to be paramount with automated technology. “It is a buyers market. What can we give with platforms to make customers happy, with this we can refine data to give them what they want. This way pop ups can come up on the site which is relevant to that user – not random popups that make no sense.”
Mr Eyal said “We have seen the real value in the hard facts with regard to AI technology. Just referring to the previous talk about deposition for the AI, I agree from the user’s needs, we need to implement what benefits the user, but exactly this is the point for the AI. If we talk about AI, as we see humans are looking for trust and human discussion and psychology involved in wealth management, but we can eventually replace the humans with machines that can react in a way that a human behind the telephone line will not.”
Will people lose their job?
Mr Hatzakis approached the subject of possible redundancies. “Some say new jobs will be created but others lost. What are your thoughts and where will automation impact on jobs?”
Mr Eyal said “Humans are losing jobs to other humans, not to tools, AI or machines. Jobs are flowing around the globe to places where it is more effective to do the job. In that sense I think community, company or culture that is a best fit to tools, and AI is just one of those tools, will get jobs. It is in the best interest of company and country to bring this to fruition.”
Mr Gat said “Decisions are being made by quants. AI is trying to solve the lack of granular information and then can serve it to users in an actionable way.”
Mr Shefer said “I think there will be less simple jobs. Account managers will evolve to manage data, build processes with higher sophistication instead of typical call center jobs.”
Concluding, Mr Ralich said “I would like to make a point here, that being from the perspective of process automation. If you have a complete system which automates risk management and platform decisions from ground up, not knowing their role, then you’re choosing the wrong provider. AI is not here yet, so if you are choosing someone who has built the all encompassing AI solution from front end to back end, you might want to look at a firm thats been around longer”