As artificial intelligence matures, it is reducing the amount of art and increasing the amount of science in illiquid bond pricing.
Markets Media speaks with Chris Bruner, managing director and head of US Credit at Tradeweb, regarding the firm’s Automated Intelligent Execution capabilities.
What does AiEX provide Tradeweb clients that they had not had before?
AiEX for Credit fully automates less complex trades, to help achieve a better price. To that end, it achieves two things: first, users can spend more time elsewhere, and on higher value trades. That helps streamline trading desks, particularly as electronic fixed income trading continues to grow globally.
Second, our new innovation called Ai-Price helps investors and traders access liquidity more efficiently and at greater scale. It creates an indicative composite price that is based on data from every corporate bond trade as well as Tradeweb trade data. That means clients can now view objective composite reference prices for more than 15,000 bonds.
To that end, AiEX helps investors and traders access liquidity more efficiently and at greater scale.
How do users access and use this capability?
The most important aspect of AiEX for Credit is that it accesses exactly the same pool of responders as a traditional all-to-all RFQ – it just automates the process. This means AiEX users reach the broadest liquidity possible from Day 1, and they can do so from their existing UI, or not at all: we also cater to those who’d prefer a zero-touch process.
It can take just a day to implement AiEX for existing Tradeweb customers, but we prefer to spend a little more time working with firms to evaluate and refine their execution rules. It’s important to understand client motivation, for example, how they trade and how they want to segregate their trading activity.
We also layer in management information systems to understand other criteria like execution timing. When you pair that with our TCA tools, it helps us (and customers) figure out which trades could benefit most from AiEX.
How does AiEX operate? Which data points does it consume?
AiEX is based on rules, which are customized by clients. Those rules can include restrictions around order acceptance, dealer selection, price negotiation, price evaluation, and counterparty compliance. Clients can adapt AiEX to suit their investment goals: some fully automate workflows, others only do so partially. It’s really up to customers.
Throughout, Tradeweb’s liquidity scores and composite price are generated by Ai-Price, which continuously updates pricing and liquidity metrics in real-time.
How long was it in development before Tradeweb took AiEX live?
It’s something we’ve been developing and refining in the US Treasury market over the years . We’ve now expanded that out into more than 25 asset classes, which is why it’s so important that we continually improve AiEX as customer needs develop and evolve.
Who is using it?
The classic case is the money manager with a significant amount of low-touch orders, that would otherwise need to be executed manually, which would be painful for the trading desk. That said, the more we extend its usage, the more use cases we seem to discover, which is a great position to be in. For example, we have found a lot of interest in automating derivatives flow. The same goes for retail aggregation firms handling multiple orders throughout the day.
How accurate is the tool and will it eventually replace evaluated pricing services?
AiEX was one of the first automated protocols to be offered by a platform, and we believe it represents one of the best available, too. Clients have achieved very high hit rates using AiEX, which is a strong indicator that the system is working as designed. We also believe, within that, that AiPrice has great potential: because we use it as a price check for thousands of automated RFQs, our pricing engine gets real-time feedback if prices appear to be out of line. Combined with all the traditional reported trades, as well as other unique Tradeweb data, that gives AiEX a real edge.
That said, we’re attuned though to what our customers want to see next. We always want to refine our understanding of where we can improve their trading process and reduce costs.
It is certainly possible that, in the future, real-time pricing engines like Ai-Price become preferred to traditional evaluated pricing services as automated trading activity grows. But there is a way to go. It’s important to acknowledge that we’re at the start of this progression, and that the considerations of liquidity, positioning, and market conditions that go into determining the prices at which participants choose to transact may never be totally mimicked by these services. Instead, they might evolve in another direction entirely, and that’s an intriguing prospect.
