Jos Evans found a way to make a trading desk erupt without a human broker negotiating the deal. In February 2020, AiX announced a bitcoin-for-dollars trade between GSR Markets and DV Chain, settled through Zero Hash. The negotiation had run through its artificial intelligence platform. In the team photograph, Evans stands near the centre holding a bottle of champagne. Around him, colleagues raise their fists. Automating a conversation had required a rather large gathering of people.
The photograph catches the appeal of his project better than a diagram of financial infrastructure could. Evans had worked inside the brokerage business, built a firm of his own, and sold it. Then he returned with software designed to take over the repetitive exchanges that made the business run. He was asking an industry to surrender some of the work he already knew how to sell.
There is a useful distinction between imagining a future trading floor and getting two counterparties through one actual transaction. Evans’s story travels between those two scales. It contains large ambitions, a funding plan that was abandoned, a chatbot with unusually demanding customers, and an eventual sale to Blockchain.com. The interesting part is how much practical work lay inside an interface that looked like an ordinary conversation.
A broker acquires a reason to change broking
Before AiX, there was Red Ops, Evans’s commodity-options brokerage. Its sale in 2012 gave him experience of building a business as well as working a market. By the time he introduced AiX, his argument came with the authority of someone familiar with the process from the inside. He considered interdealer broking expensive, slow and opaque. Those complaints became the brief for his next company.
His proposed remedy was deliberately easy for a trader to recognise: a chatbot. Clients would ask for prices and negotiate using language, while software handled the work behind the exchange. In the 2017 launch announcement, AiX proposed connecting cryptocurrency and traditional financial markets through this interface. The ambition extended to recording the decisions behind trades in an auditable form. Familiar conversation at the front; a substantial engineering project behind it.
Evans also approached technology as an investor. An early connection was Rainbird, the decision-making software company founded by James Duez. He had personally invested in the business. Seeing what its technology could do helped lead him towards AiX. His route into automation therefore ran through a company he had backed, as well as through the brokerage he had operated.
Early development drew on relationships from his previous working life. Counterparties let the team use instant-message transcripts to teach the system how trading conversations worked. That detail makes the project unusually concrete. Financial language comes with local habits and specialist shorthand. The software had to learn the conversation people actually conducted, complete with its awkwardness, before it could reliably participate.
The price of an ambitious prediction
Evans did not introduce AiX with a modest forecast. In June 2018, he set out targets of five to ten per cent market share in the first year and 75 per cent of interdealer-broker revenues after five. Those were his ambitions. They convey the scale of the challenge he wanted to mount, rather than a measure of what the company subsequently achieved.
The underlying commercial argument concerned commissions. If software could do more of the work involved in finding and executing a trade, AiX believed it could charge less. Evans talked about reductions ranging from 50 to 95 per cent. The spread itself is revealing: different markets would offer different opportunities. A costly process can make an inviting target, but replacing it still means persuading clients to entrust their transactions to the replacement.
There was an earlier technical milestone, too. In June 2018, AiX announced an AI-brokered cryptocurrency trade between Rockwell Capital and TLDR Capital. This preceded the 2020 platform launch and its GSR-DV Chain transaction. Treating these as separate moments makes the development easier to understand: a working demonstration could establish a possibility, while a launched service needed a fuller operating system around it.
The dates also give the story some resistance to the instant-invention myth. AiX began in 2017; announcements and demonstrations followed; the launch arrived in 2020. A simple-looking product can have a lengthy apprenticeship. In a market where the details of a transaction matter, the distance between an idea and a dependable service has to be travelled one instruction at a time.
Money was only one of the moving parts
AiX’s early fundraising plan belonged to the token-sale era. In 2017, the company was preparing an initial coin offering with a proposed $200 million target. The sale did not proceed. Evans later explained that the company had been unable to obtain explicit British regulatory approval and decided the future risk was too great. The abandoned plan is part of the business’s history, rather than money it raised.
By February 2020, the company reported $8 million raised through seed and Series A funding. It had 35 staff, including 26 programmers and engineers. The team photograph’s crowd begins to make sense. The company that wanted to automate a broker’s work had become, in substantial part, a software-building operation. Settlement proved especially difficult: a negotiated agreement still needed a dependable way to complete the exchange of assets.
“You can swear at it, but the system won’t swear at you.”
Jos Evans, on the AiX chatbot, 2020
The joke gives the product a small advantage in office manners. It also brings Evans’s argument down to human scale. The interface was meant to fit into the way traders already communicated. A client could be impatient, terse or colloquial; the service still needed to carry the transaction forward. Getting the tone right was pleasant. Getting the instruction right was the work.
A conversation that ends with settlement
The February 2020 launch tied several functions together. AiX described a service handling price discovery, negotiation, execution and settlement, with traders interacting through the chatbot. GSR Markets and DV Chain supplied the counterparties for the first announced negotiated BTC/USD trade on the launched platform. Zero Hash provided the settlement infrastructure. Blockfills and Trigon Trading were also named among onboarded participants.
That sequence matters because each stage answers a different question. What price is available? Will the parties agree? Has the trade been executed? How will the exchange be completed? A convincing answer to the first question leaves the others waiting. Evans’s product proposition depended on joining them, so that the convenience of a chat interface extended beyond a useful price quotation.
- 01 Discover a price
- 02 Negotiate terms
- 03 Execute the trade
- 04 Settle the assets
Four connected stages in AiX’s announced institutional trading service.
At a 2019 industry discussion, Evans had already identified asset security as a barrier to institutional participation in crypto. The execution process remained fragmented, with conversations conducted through chat applications and settlement handled manually. His concern was practical: institutions needed infrastructure that made participation workable. An attractive market still had to offer a usable way in and a dependable way through.
That perspective helps explain why his work kept returning to connections between systems. A trader, a market maker and a custodian all had parts to play. Improving the conversation alone would leave the surrounding process intact. The opportunity Evans pursued lay in making those separate parts cooperate with less manual intervention. The customer could continue typing while more of the machinery behind the message became connected.
What the software should ask back
In a May 2020 interview, Evans explained a difficulty that sounds almost absurd until money is involved: general-purpose language systems struggled to distinguish the bid, the offer and the volume. AiX added proprietary elements to address that problem. It also confirmed instructions back to clients, allowing opportunities to change or cancel before execution. The system’s ability to ask a question was part of its usefulness.
His view of the human broker had become more nuanced than the early replacement rhetoric suggested. He argued that automation could leave people opportunities in sophisticated analysis and finding ideas for clients. The repetitive process could change while the relationship acquired different work. That was an aspiration for the profession, with the commercial and technical argument meeting at a familiar question: what should a person spend their time doing?
Another company buys the conversation
In December 2020, AlgoTrader announced a partnership with AiX. The arrangement aimed to give institutional clients access to its automated OTC service and connect traders across markets. Evans welcomed the fit between the two companies’ technologies. Distribution had become another part of the undertaking: the system needed routes into the institutions whose workflows it hoped to improve.
The partnership also placed AiX beside an existing provider of algorithmic trading and execution software. Its chat interface was one entrance to a wider set of trading functions. For Evans, the business had moved from persuading people that a broker could be automated towards connecting that capability with other infrastructure. The conversation had to reach the right desks.
On 4 May 2021, Blockchain.com announced that it had acquired AiX’s business and would welcome its team. The announcement highlighted the negotiation and matching engine, Evans’s derivatives experience and chief operating officer Taylor Cable’s background at Moore Capital. Blockchain.com wanted to expand its institutional products. AiX brought technology and people who understood the market it was serving.
Evans had now built and exited two businesses. One sold brokerage services; the other built a way to automate their delivery. The champagne photograph sits between those chapters, at a moment when the system had completed a real trade and its builders could enjoy the result. Around the chatbot, people had plenty left to say.