By Ray Wong, Director of Product - AI & Data

 

The learning curve is still steep

I joined Pirum just over six weeks ago, and my first week involved a fair amount of meetings where I had to ask for the meanings of the frequently occurring acronyms. CSDR. RQV. SSI. ISITC. I was also learning the difference between a recall and a return thanks to very patient explanations from internal subject matter experts, many of whom had been in the industry for twenty years.

I am writing today very much still on this learning curve, but with a clarity that there’s so much value in coming to securities finance from building and deploying AI systems elsewhere.

I can see something that people who have been in this industry for years are often too close to see clearly. That is why I can confidently say that Pirum is unusually ready for this moment.

What “ready” actually means

Irrespective of the industry, the three key things that a firm needs to be “AI-ready” are infrastructure, people, and platform. And all three have to be ready at once. One or two out of three will not suffice – just imagine a tricycle with a wheel missing.

Infrastructure, first. Pirum’s stack is cloud-native, not simply cloud-hosted. When I want to test a new model or deploy an agent framework, that requires a conversation and a sandbox, with anonymisation, segregation and other data protection already in place automatically. In many firms with legacy architecture, that would mean a six-month infrastructure project, and reaching production data would likely take the best part of a year.

People, second. I expected some resistance to fully fledged AI adoption, as that’s usually the case when a new technology turns up and people wonder what it means for their role. I have not found any of that here. Engineering, product, sales… everyone I have spoken with is working on using AI to upgrade what they can offer clients. They are doing so thoughtfully and within our robust AI governance and risk framework.

Platform, third. A model is only as good as what it learns from. Pirum is building on over twenty-five years of data, workflows, and counterparty relationships that are running in production day in day out. While many firms trying to “do AI” start from data sets that are incomplete or inconsistent, Pirum’s data is alive, real-time and uniquely reflecting how hundreds of financial institutions are behaving. It would be very difficult indeed to build a training set like that in a year or two – not one drawn from a network processing US$6.5 trillion in daily transactions across 150+ institutions worldwide.

What this means for securities finance

Those three things – infrastructure, people, platform – in a mature state enable a company to be ready to add AI to their product. They do not, however, enable a company to become “AI-ready” in our industry. For that, we need to add two more ingredients: deep securities finance knowledge and clear, client-validated use cases (as opposed to vague “let’s make an agent” ambitions).

Knowing that a recall and a return are not the same thing, or what happens when a corporate action lands on an open loan, is naturally the bread and butter of a securities finance operator. But it is not something that a model learns from a clean dataset. The model requires tech-savvy experts driving product development with the benefit of deep securities finance know-how.

The real-world use case also has to be specific – which is where the clients behind the experts come in.

Clients know their pain points best and enable an AI developer to zero in on the most salient problems. A settlement risk signal a desk can act on, or a benchmarking figure a client can use in a negotiation, for example. A vague “AI for securities finance” ambition is the fastest way to burn through a budget and end up with a solution looking for a problem.

I have not, in my career so far, walked into a place with all five ingredients sitting there waiting on day one. Three of them get you readiness in general. The other two are what make it real in this industry specifically. That is what caught my attention about Pirum before I had signed anything. It is still what strikes me now, six weeks in. And it is truly exciting.

If you want to talk through what AI-readiness looks like for your own operation, get in touch.

Also included in the Quarterly Panorama:

  •  Ben Challice, CEO, on Vision 2030, and why scale and connectivity are now table stakes, with the real advantage coming from turning network-wide data into intelligence and automation. 

  • Zoë Balkwell on pre-trade intelligence, and why the real value of AI is surfacing opportunities a desk would never have known to look for.

  • Renato Lima, Chief Commercial Officer, on the build trap, and why in-house AI starts at near zero when a coding agent cannot replicate 25 years of counterparty trust.

The Panorama also includes a regulatory and events calendar.