To what extent will the future of e-FX distribution belong to APIs rather than screens?
Having worked in FX markets for over 25 years, I’ve seen distribution evolve through several generations of traders and technology, but I think the next couple of years could illustrate one of the biggest shifts I’ve seen. Putting it into context, I’d compare what’s coming to how the internet transformed the way we worked in the late ‘90s.
Routine FX will increasingly happen in the background, initiated by another system as opposed to somebody manually logging onto a trading platform. This programmatic shift is set to be the most transformative change. In our recent survey of banks and corporate treasurers, corporates expected the combined share of embedded FX and user-triggered APIs to grow from 18% in their current execution mix to 42% in five years.
Screens won’t disappear, but their role will certainly change more than we’ve seen in the last decade. They will remain important for exception management and complex trades where human judgment is key. I think it’s the same for voice. For as long as I can remember, the industry said, ‘voice is dead’. But voice didn’t die, because there are certain workflows and reasons where a direct conversation between the bank and treasury team is an important conversation to have.
Integral’s research suggests that corporates expect embedded FX to more than double its share as an execution channel over the next five years. What needs to change, technologically, operationally or culturally, for that shift to happen at scale?
Based on what we’ve heard from discussions with clients, I’d highlight three trends that stand out as particularly influential: connectivity becoming easier, infrastructure becoming more interoperable, and on the cultural side, institutions becoming more comfortable with programmatic FX in practice.
In the past, the real driver of change wasn’t necessarily new or more advanced technology. It was the ease of adoption. Taking the use-case of embedded FX, you’re not asking the client to change their behaviour by going to another screen or changing their processes. Your clients don’t need to go somewhere else to trade FX. FX comes to them. This is why its adoption has a greater chance of reaching its predicted scale. Our research shows embedded FX rising from 12% to 26% of corporates’ execution mix over the next five years. The fact that 82% of the banks and other financial institutions we surveyed agree that embedded FX will play a significantly greater role in their distribution strategy reflects what we’re seeing anecdotally about the consistency of the direction of travel.
How can banks preserve differentiation when FX becomes increasingly invisible to the end user?
FX becoming less visible to the end user doesn’t mean that the bank’s ability to show differentiation disappears. Banks can still generate competitive advantage through the quality of pricing, liquidity, execution, and risk management, alongside customer service and relationship management.
At Integral, we’ve always focused on equipping institutions with infrastructure to create their own differentiated, tailored service through customizable technology stacks and trading interfaces, as opposed to taking a one-size-fits-all approach for every client. With strong indications that FX is becoming increasingly embedded, that level of configurability and focus on bespoke service are going to become even more valuable.
How can banks centralize pricing, liquidity and risk management while FX distribution becomes decentralized across multiple client applications and workflows?
It’s not a situation where the two are mutually exclusive. Whether a client accesses FX through an ERP, treasury management system, or payments platform, every channel can connect back to common pricing, liquidity, and risk infrastructure. This means banks gain a consolidated view of activity and risk, while clients access FX through whichever workflow they prefer. What’s key going forward is to avoid creating a new system every time a new distribution channel emerges.
Does embedded FX need to solve different problems across the market, given the primary barriers to automation vary for firms of different sizes?
Absolutely. One of the most interesting findings from our recent survey is that the primary barrier to automation differs significantly across different parts of the market. For large corporates, for example, 56% stated that the main obstacle was their own legacy architecture. For mid-market firms, the most cited response was banks’ API complexity and lack of standardization.
The simple suggestion to build more APIs isn’t really going to address the situation at hand. And as technology spending grows, particularly with the rise of AI, institutions need to be disciplined in looking at what solutions are going to make the biggest difference within their unique business context and avoid automatically assuming that what works for others will work for them.
If FX execution takes place inside a treasury management system (TMS), ERP platform or payments workflow, how do banks ensure they can still deliver tailored pricing to clients at scale?
If a request arrives via a TMS or ERP rather than a dealer screen, the bank can still account for the client relationship, transaction size, currency pair, and risk position. In fact, centralization makes personalization easier because, when a bank has a more consolidated view of its FX flows and risk exposures, it has more information on which to base pricing decisions. The API, or ultimately the AI agent using it, is simply another way of connecting the client’s workflow with the bank’s underlying FX capabilities.
Automation doesn’t remove the obligation for best execution, of course, so banks offering embedded FX must deliver it through competitive pricing and then prove it, with TCA.
Could the competitive battleground in e-FX shift away from trading interface aesthetics and towards the orchestration layer that underpins it?
A slick, intuitive interface obviously remains important for user experience, but an ERP system or AI agent doesn’t care what the screen looks like. That’s not to say style and substance don’t coincide, particularly when traders need an accurate visualization of prices, liquidity and risk to make quick decisions. But as more execution happens programmatically, the orchestration layer becomes a fundamental source of differentiation.
What needs to happen for agentic AI to play a meaningful role in FX execution?
Agentic AI will only play a meaningful role in FX execution once the underlying technology stack and operating model are ready to implement it. Banks need to put in place capabilities for real-time data, interoperable APIs and, crucially, connectivity that supports bidirectional workflows rather than simply pulling information from a system.
Another key prerequisite to meaningful adoption is trust and governance. Our survey found that only 8% of corporates are actively piloting AI in treasury today, while 67% remain in the monitoring or conceptual phase. Near-term adoption will primarily be focused on deploying agentic AI within clearly defined guardrails, with humans retaining control over strategy, limits, and exceptions. If that balance can be achieved, then agentic AI integration will move from an interesting concept to a practical part of day-to-day FX execution workflows.
AI is really powerful when it comes to looking through huge amounts of data and drawing insights. But for certain workflows, deterministic technology must be used. The last thing you need when you’re trying to buy a hundred million euro is for an agent to go off and sell it instead, only for the prompt to come back, “Oh, I’m sorry. What would you like me to do next?” If you actually had a trader do that, you’d probably get fired. You can’t fire an AI agent. So, there must be a clear distinction on what’s managed by a GenAI agent versus other forms of automation.
How do execution governance and best-execution oversight need to evolve as more FX transactions are triggered automatically as part of another workflow rather than by a trader clicking a price?
Execution governance needs to focus on supervising the rules behind automated decisions. Firms need to know why a transaction was triggered, the data that informed it, what liquidity was available at that time, and when human intervention occurred. This level of transparency will become even more critical as agentic AI becomes more deeply integrated into trading systems. In a similar light, best-execution oversight becomes about assessing the decision-making framework as well as reviewing the individual trades. In short, internal auditability and compliance policies, alongside the scope of regulation, will have to advance in step with the infrastructure they oversee.