THE GIST
Our analysts just identified a stock with the potential to be the next Nvidia. Tell us how you invest and we’ll show you why it’s our #1 pick. Tap here.
ING is using AI to build electronic trading systems in hours rather than weeks. That sounds like a productivity story. It’s also a risk story, a competitive story, and a glimpse at where the entire banking industry is heading whether it is ready or not.
WHAT HAPPENED
ING’s global head of electronic trading has confirmed the Dutch bank is using vibe coding, the practice of describing what you want to a large language model and having it write the code, to build trading infrastructure for its foreign exchange and credit businesses. Recent applications include an entire credit e-trading system and real-time analytics dashboards showing pricing, incoming trades, and performance metrics. Work that previously took a team of developers weeks is now being completed in hours.
The bank’s preferred external model is Anthropic’s Claude. ING credits an in-house AI currency pricing model with an initial 50% increase in large-ticket trades, a number that points to meaningful commercial impact rather than experimental tinkering. The electronic trading team numbers around ten people, a fraction of the headcount at larger rivals, which is precisely why the bank has leaned into the approach: it can’t compete on raw resource deployment, so it is competing on how creatively it deploys what it has.
ING separately announced it will cut around 1,250 operations roles globally in 2026 as part of a broader response to AI and digitalization. The head of electronic trading was candid that AI is reshaping what kind of people trading floors want to hire, with the future workforce expected to center on fewer, more technically sophisticated employees supervising complex systems.
The week also produced a high-profile moment of blowback on the human cost of that transition. Standard Chartered CEO Bill Winters apologized on LinkedIn after describing staff set to lose jobs to AI as lower-value human capital during a presentation about plans to cut around 7,800 back-office roles by 2030, or roughly 15% of the bank’s operations workforce. The apology was widely viewed as inadequate, with commenters noting that the clarifications he offered largely restated the original sentiment.
WHY IT MATTERS
ING’s vibe coding story is being told as an efficiency win, and it is, but that framing undersells what is actually happening. A major bank is now deploying AI-generated code directly into trading systems that handle tens of billions of dollars of activity per day. That isn’t a prototype. It’s production infrastructure, built faster and with less human review than anything that came before it.
One stock. Nvidia-level potential. 30M+ investors trust Moby to find it first. Get the pick. Tap here.
The competitive logic is straightforward. Larger banks can throw capital and headcount at technology problems. ING cannot, so it is using AI to close the gap. The 50% increase in large-ticket FX trades attributed to the AI pricing model suggests this is working. If a ten-person team can build and deploy systems that previously required far larger engineering organizations, the traditional advantage of scale in electronic trading starts to erode. That is a genuine threat to the firms that have been winning on the basis of technological investment over the past decade.
The risk dimension is less comfortable but equally important. AI-generated code running in a live trading environment introduces failure modes that do not exist in human-written systems. Models can hallucinate logic errors. Edge cases that a senior developer would catch may not surface until they cause a real problem. Code that passes testing in normal market conditions may behave unexpectedly in stress scenarios. The financial industry has spent decades building controls around human-written trading code, but the risk frameworks for AI-generated code are still being written.
There is also a longer-term question about technical debt. AI-assisted code tends to accumulate structural weaknesses at a faster rate than conventionally written software, creating maintenance burdens that are not immediately visible but compound over time. A trading system built quickly through vibe coding may also require significantly more intervention to maintain, debug, and upgrade, adding inference workload and engineering overhead that offsets some of the initial efficiency gain.
The Standard Chartered situation is a useful counterpoint to ING’s more forward-looking framing. The efficiency gains from AI in banking are real, but so is the displacement of tens of thousands of workers, and the language that executives use to describe that displacement matters enormously. Winters used the term lower-value human capital to describe people losing their jobs, and the response suggested the industry has not yet found a way to talk honestly about what is happening without causing justifiable offense. The apology did not change the substance of the announcement, just the framing.
WHAT’S NEXT
The question for the industry is how quickly vibe coding moves from the early adopters to standard practice, and whether the risk frameworks keep pace with the deployment. ING’s head of electronic trading predicted the approach would become widespread in banking within a year.
If that timeline is accurate, the combination of persistent AI inference load, accumulating technical debt, and compressed development cycles will create a new category of operational risk in financial markets that regulators are only beginning to consider.
Disclaimer : This story is auto aggregated by a computer programme and has not been created or edited by DOWNTHENEWS. Publisher: finance.yahoo.com
