Earlier this month, I contributed an article to FX Markets examining a simple but uncomfortable idea: risk in electronic FX has not disappeared. It has moved.
What was once explicit and human is now increasingly embedded in systems, parameters, and automated decision logic. This post does not restate that argument. Instead, it looks at what follows from it.
Because the more important question is not whether risk has migrated, but whether institutions are still equipped to recognise it.
Risk has become harder to see, not smaller
Electronic FX is often described as operationally mature. Execution is fast. Routing is sophisticated. Internalisation rates are high. On most dashboards, the machine appears stable.
That appearance is misleading.
As pricing, skewing, hedging, and execution controls have been automated, risk has become distributed across models rather than concentrated in desks or portfolios. The result is not lower risk, but less observable risk.
When outcomes deteriorate, the post-mortem tends to sound familiar:
- No single control failed
- Each component behaved as designed
- The loss emerged from interaction effects
This is rarely a technology failure. It is a visibility failure.
Automation does not remove judgement. It encodes it
One of the most persistent misconceptions in eFX is that automation removes subjectivity. It does not. It simply relocates it.
Every pricing engine expresses judgement, whether explicitly acknowledged or not:
- How much inventory is acceptable
- How aggressively adverse selection is tolerated
- How the system should behave when liquidity thins
- When responsiveness should slow or stop
These judgements used to sit with traders. Today they sit in configuration files, thresholds, and fallback logic. Because they are expressed as parameters rather than opinions, they are often treated as neutral infrastructure.
They are not.
They are risk decisions that have been made once and then left to run.
Where issues actually surface
In practice, stress rarely arrives where firms expect it. The most damaging problems are seldom headline market events. They are slow-burn effects that build quietly inside the system:
- Internalisation ratios drifting higher without capital context
- Hedging logic lagging regime change
- Skew controls interacting non-linearly with client behaviour
- Risk metrics that remain benign until they suddenly are not
By the time the issue is visible at the P&L level, the causal chain is already obscured. Responsibility is diffuse. The machine did what it was told to do.
This is not a quant problem
These issues are often framed as quantitative or technological. That framing is incomplete.
The real question for senior decision-makers is simpler and more uncomfortable:
Who is accountable for the behaviour of the system as a whole?
Not the model in isolation. Not the venue choice. Not the liquidity mix. The composite behaviour that emerges when all of these interact.
When judgement is embedded into systems without ongoing scrutiny, organisations tend to optimise locally and fragilise globally. Controls become internally consistent but externally brittle.
The practical implication
Most firms do not need more dashboards. They need clearer narratives about how risk is expressed across their infrastructure.
In practice, that usually means:
- Making explicit where judgement has been displaced into automation
- Stress-testing interactions rather than individual components
- Treating configuration choices as risk decisions, not technical settings
- Reintroducing ownership for system behaviour, not just system uptime
This is less about rebuilding platforms and more about reframing how they are understood.
Closing thought
The FX Markets article was intentionally observational. This piece is intentionally operational.
If risk has moved into the machine, then oversight has to move with it. Otherwise, firms are not actively managing risk. They are inheriting it.
