Joining me to discuss what is happening in financial services is Conor Twomey, founder and CEO of AI One. Conor, it's a pleasure to have you with us today.
Thanks Johnny. Delighted to be here.
Let's start with AI One. What are you doing in financial services?
My co-founder Fergus Keane and I come from a deep background of data analytics in capital markets. Our contention was that as organisations tried to absorb this new technology, getting accurate, timely, and confident responses or actions was going to be very difficult. So we invented a lightweight software layer that underpins strategic AI platforms, specifically to improve the accuracy, repeatability, and token economics of either actions or answers.
Banks and insurers have had hundreds of AI pilots. You're at the front lines of this every day. Why do so many get stuck, and what value are firms actually getting?
There's a lot of sizzle out there. A lot of noise. A lot of people trying to create forward-looking statements about value generation for valuation purposes. What we find is that the people doing this well are showing it in their financial profile — not talking about dozens or hundreds of pilots, but showing how they are using this technology to grow their top line through new value-added products and services, accelerating their roadmap, or moving more aggressively into adjacent markets. The best of the best are simultaneously raising their top line while reducing their cost base. That jaws effect is the real signal. There's nowhere to hide. It needs to show up in the P&L or it's noise.
A lot of people think this is just a data challenge — get the data right and everything else follows. What's your view?
We call that a false summit. There are five or six false summits that convince organisations they are almost at the very top — and then the clouds clear and they realise they're still at base camp. Having systems you can actually trust, that provide accurate and timely responses, is non-trivial.
Where are institutions actually investing in AI right now — and what are you seeing across the Americas versus MENA?
Within the Americas, what we see on the East Coast is a very myopic view — does this save me time, save me money, or reduce risk? It's about improving the world as it exists today. On the West Coast, the view is far more forward-looking — if the world moves in a certain direction, how do we position to capitalise on it? West Coast is building for tomorrow. East Coast is building for today.
From a geographical perspective, demand is everywhere — this technology needs to permeate every organisation, every scale, every sector, every geography. We are about 5% of the way there. One of the core challenges is not just technology — it's organisational structure. The firms moving fastest and showing results in their P&L have cracked a hub-and-spoke model: a centralised centre of excellence setting the standards and surveying the vendor landscape, combined with genuine autonomy at the line-of-business level. Firms doing it badly are either stuck in analysis paralysis with everything centralised — death by consensus — or they allow everything to happen at the line-of-business level and get chaos. In the Middle East, we see a lot of centralised decision-making, similar to the East Coast. Speed and experimentation can suffer as a result.
What does AI running daily operations actually look like from a large financial institution's perspective?
We've had roughly 20 years of labour arbitrage — middle and back office work migrating to locations like Chennai, Pune, and Costa Rica. Now people are realising there's a software-based approach to that same activity. We call it high-complexity, low-judgement work. High complexity because anything low or medium complexity has already been automated with if-else decision logic and robotic process automation over the last 15 years. Low judgement because large language models are not in a position to replicate a subject matter expert. They haven't spent 20 years at a desk. They've been trained on Reddit and the internet. There is no substitute for that lived experience and that pressure of real decision-making.
Do heavily regulated institutions actually trust AI with this work?
You cannot outsource regulatory work to AI. AI is inherently probabilistic. If I asked you to complete the nursery rhyme "Mary had a little" — using probabilistic technology, 999 times out of a thousand it will say lamb. But one time out of a thousand it will say lasagne, because somebody on Reddit talked about the little lasagne they had for dinner one evening. That is the nature of probabilistic technology. The real skill is knowing when to use probabilistic AI to support decision-making, and when to use good old deterministic logic — same inputs, same outputs — which we have relied on for 60 years. Being able to weaponise both is what organisations are really grappling with right now. Don't get AI to do the work. Get AI to give you operational leverage. Nobody wants AI. They want the ability to do more with the resources they have — to look and feel like an army. That is the North Star for all of these organisations, including their risk departments.
Conor, thank you so much for joining us. It's genuinely fascinating to hear your perspective on how AI is reshaping these institutions.
Thank you Johnny.
