A new company in London, A.I. Exchange, or AiX, seeks to disrupt the global interdealer-broker oligopoly by using technology to vastly cut costs – and commissions.
The fledgling company plans to raise $200 million via an initial coin offering (ICO) this October. If successful, the money will go to hiring management teams in Singapore and New York, as well as in London, and assure potential clients that it can meet collateral requirements.
It has also just hired J.C. Oliver, previously the global head of innovation at Microsoft in London, as an advisor.
The founder of AiX, Jos Evans, and his team are nearing completion on an interdealer-brokerage product based on artificial intelligence and blockchain. They plan to then seek brokerage licenses in the U.K., Singapore and the U.S., with the goal of launching in all markets and asset classes, in all three trading centers, simultaneously – perhaps by autumn 2018. Evans was previously founder of Red Ops, a commodities-focused IDB that he sold to a competitor.
He expects AiX’s platform will be able to reduce headcount costs, and by extension commissions, by up to 90% for the most exotic markets and by 50% for standardized products.
Breaking the oligopoly
On the outside, otherwise, the firm would offer an orthodox experience for clients such as commodity trading firms, hedge funds and investment banks. But the radical cut in price will attract liquidity, Evans hopes, thereby letting AiX break into a small market of large IDBs that dominate the market.
The top three IDBs – BGC Partners, T.P. ICAP and Tradition – together commanded revenues of $4.8 billion in 2016, according to AiX.
IDB execution is mostly just monkey business
DigFin spoke with a retired IDB executive who agreed the industry is ripe for disruption. IDBs have faced pressure on margins by consolidating into an oligopoly, instead of innovating, he said. They serve a useful purpose by tailoring trades, giving institutions bespoke views on risk.
But to maintain profitability, this former trader said, IDBs pitch increasingly opaque, complex and expensive ideas, when many institutions could probably get a similar exposure on an underlying security by simply selling an option. As a result, these firms have largely resisted the trend toward electronic trading that characterizes an investment bank’s equities or commodities desk, for example.
Evans is more blunt. “IDB execution is mostly just monkey business,” he said, noting that brokers seek prices for clients by blasting requests out via instant messaging or tap banks’ opaque dark pools. “Traders [clients] assume they’re getting screwed.”
Transparency and trust
But a fully automated system, in theory, should improve trust. The ability to be transparent about how a computer weighs probabilities to make a decision (or decide against a course of action) is something that both regulators and clients should welcome.
AiX is using technology that it says is auditable, developed by Rainbird, a software company founded by software entrepreneur James Duez in which Evans is a personal investor. In fact, it was this personal investment that led to AiX, once Evans realized what this A.I. provider was capable of. Rainbird currently works with customers such as Barclays to detect fraud.
Rainbird is also developing a shared distributed ledger for AiX, which it will use for recordkeeping and auditing (but not for execution). Evans acknowledges blockchain isn’t required for keeping track of trades, but says using it will enable the A.I. to learn faster. Permissions on the ledger will be limited to AiX’s senior managers, so outsiders won’t see transaction histories other than their own, although Evans says there could be a role to allow regulators to have some level of access.
To teach the machine how to construct and price a trade offer, AiX has been feeding it transcripts from instant messages. Evans says he obtained these through his personal relationships with a few counterparties that are likely to become AiX’s first clients; they are letting his machine read their conversations with other IDBs in order to train it, so they can begin to use it.
A new jargon?
“We’re starting to build the product,” Evans told DigFin. “We’ll test it and run transactions. Then we can show it to regulators, so they can help us understand all the risks we need to address, and understand how it works. Then we can get the necessary licenses.”
But will the chatbots understand it?…We’re proving the taxonomy
The hardest point about building the product is idiosyncrasies in the lingo. Finance is notorious for its jargon, and people on different desks or in different countries use different words that mean the same thing. “But will the chatbot understand it?” notes James Tabor, a cryptocurrency advisor and head of AiX’s marketing. “We’re proving the taxonomy.” Rainbird’s A.I., which is expected to look more like a conversation between two people than filling in an order (think Amazon’s Alexa), will also confirm the details before executing an order, to be sure the terms are clear.
Assuming the A.I. is good enough and the chatbots are sufficiently trained – and that the company can get licensed in all three markets within a short period of time – it will then be time to attempt to take on the oligopoly – whose members aren’t sitting still, waiting to be disrupted.
But Evans says his competitors are using innovation to improve their existing business model, not upending it, and he says he can run a book of business as big as BCG’s with only 10% of the headcount.