MiFID’s fixed income perimeter was not a mistake from end to end.

One of MiFID’s more sensible moves was the creation of the OTF category. It acknowledged that non-equity markets, especially bonds and derivatives, often rely on discretion, negotiation and hybrid workflows that do not fit neatly inside a purely non-discretionary MTF model. That was a better regulatory fit, even if it did not lead to a clean market shift from MTFs to OTFs.

MiFID also got one broader point right. European bond markets needed a stronger transparency framework. Pre-MiFID II, bond transparency was fragmented, patchy and commercially gated. However flawed the implementation turned out to be, MiFID at least established that post-trade visibility in fixed income should be a regulatory objective rather than a vendor side-effect. The FCA’s later reforms do not reject that objective. They are an attempt to fix a regime that was too complicated and too weak in practice.

And finally, regulators were right to worry about venue-like behaviour appearing outside venue rules. ESMA’s work on the trading venue perimeter was a recognition that systems can be dressed up as workflow tools while performing functions that look suspiciously like organised trading. That perimeter discipline still matters. A market cannot function well if every quasi-venue can claim to be nothing more than clever middleware.

Where MiFID struggled was not in identifying the need for transparency or discipline. It struggled in identifying the right unit of regulation.

Fixed income is not just a venue market. It is a workflow market.

The economically meaningful parts of execution often sit across dealer selection, negotiation, protocol choice, timing, routing, pre-trade analytics, information leakage control and exception handling. The trade print is only the end of that chain. MiFID’s architecture, by contrast, focused heavily on named containers: RM, MTF, OTF, SI, APA and so on.

That led to three major problems.

The first was the SI regime in non-equities. In the UK at least, that experiment has now effectively been written off. The FCA said respondents agreed that the SI regime for bonds and derivatives no longer contributes meaningfully to transparency or market integrity, and removed it from 1 December 2025. That is not a trim around the edges. It is a sign that one of MiFID’s key classification tools failed to earn its keep in fixed income.

The second was the quality of transparency itself. MiFID generated reporting obligations, but not necessarily usable market truth. The bond market ended up with too much data sludge, too much fragmentation and too little genuinely actionable post-trade transparency. That is precisely why the consolidated tape matters now. The tape is an admission that a patchwork of disclosures and reporting endpoints was never enough.

The third, and more important in strategic terms, is that MiFID’s perimeter logic helped entrench a venue-centric operating model.

Instead of the buy side owning the execution brain in its own stack, much of the electronic bond market evolved so that workflow, configuration, automation and intelligence became embedded in venue-operated environments. Bloomberg, Tradeweb and MarketAxess did not just become execution destinations. They became workflow environments, complete with their own interfaces, settings, protocol logic and increasingly sophisticated tooling.

That may have accelerated digitisation. But it also fragmented the buy side’s execution intelligence across external platforms.

This is where the debate becomes more interesting.

A venue should be where I execute, not where my execution logic lives.

In equities, if I want to trade on the LSE, I connect programmatically. In listed derivatives, if I want to trade on Eurex, I connect programmatically. My rules, my controls, my analytics and my automation sit in my own stack. The venue is an execution endpoint.

Why should fixed income be different?

Why should a buy-side firm need to think inside a venue UI at all?

The usual defence is that fixed income is more complex, more fragmented and more relationship-driven. Fine. But that is an argument for smarter buy-side workflow, not permanent dependence on venue-native environments.

Once intelligence is fragmented across venue-specific interfaces and rules engines, the buy side pays several times over.

It loses strategic control because its most valuable logic no longer sits centrally in its own EMS or execution fabric. It loses efficiency because the same intent must be translated into multiple external workflows. It loses governance because controls, audit trails and exceptions become fragmented across different systems. And it takes on more operational risk because each venue comes with its own UI conventions, configuration models, support processes, release cycles and embedded assumptions.

That is not just inconvenient. It is structurally inferior operating model design.

A firm should not have to maintain a patchwork of mini execution operating systems, each owned by a venue, just to trade electronically in one asset class.

This is not only an operating model issue. It is also a competition issue.

If the venue owns the workflow, the venue captures the learning.

That matters because the future edge in fixed income is increasingly not just where the trade prints, but how the workflow is orchestrated beforehand. Dealer selection. Protocol selection. Timing. RFQ sequencing. Leakage control. Adaptive automation. Decision support. Cross-channel routing. These are exactly the areas where AI and advanced execution intelligence are likely to create the most value.

If the buy side’s own systems are reduced to little more than pipes that send orders into venue logic, then buy-side innovation becomes cramped from the outset.

The venue operator gets to build the richer workflow layer. The buy side gets to configure around the edges.

That is not a neutral outcome. That is a market structure in which the innovation advantage compounds inside the platforms that already sit closest to liquidity.

And yes, this has an information dimension too. In fixed income, showing your hand too early is not a side issue. It is a core execution risk. If the market design nudges firms towards externalising intent through venue-centric workflows when they should be able to observe liquidity, analyse options and stage decisions inside their own stack first, then the perimeter has not just shaped compliance. It has shaped bargaining power.

AI makes the flaw more obvious.

The most valuable AI use cases in fixed income are not simply about auto-executing on a venue. They sit earlier in the chain from predicting which dealers are genuinely in the bond, selecting the best protocol for the instrument and size, sequencing outreach, adapting RFQ waves, reducing signalling risk, identifying when to stay bilateral, and learning from quote quality and responsiveness over time.

Those models work best when the firm has a central decision layer, central data, central feedback loops and central governance.

Venue-centric workflow pushes in the opposite direction. It fragments the data, fragments the controls and fragments the learning environment. Each venue becomes its own little kingdom of UI, logic and exceptions. AI does not thrive in feudalism.

If MiFID’s fixed income perimeter helped encourage a world in which workflow intelligence is embedded inside venue-operated environments, then the same structure now risks constraining the next phase of AI-native execution.

This is the moment where the policy conversation should move on.

For years, regulators could argue that if fixed income workflow remained too dispersed and too opaque, then keeping a tighter perimeter around execution environments was part of how they preserved visibility.

That case weakens once the tape becomes real.

The exact details of implementation still matter, of course. But the direction of travel is now clear in both jurisdictions.

That means policymakers have a better answer to the transparency question.

If the tape becomes the transparency backbone, then not every meaningful pre-trade or workflow function needs to be forced into a venue-owned environment in order for the market to remain visible. Transparency can move to the consolidated post-trade layer, while workflow intelligence can move back towards participant-owned stacks.

That is the real case for controlled de-perimeterisation.

Not less transparency. Better placement of transparency.

The argument is not that venues should disappear. Nor is it that workflow tools offered by venues have no value. Some are genuinely sophisticated and useful.

The argument is that they should be optional.

In a mature market structure, venues should expose executable protocols, liquidity access and data services cleanly through APIs. Market participants should be able to consume those services through their own workflow environments, with their own rules engines, their own AI, their own controls and their own audit architecture. Native venue tools should still exist, but as optional front ends rather than compulsory centres of gravity.

That is not a radical proposition. It is how mature electronic market design already works in other asset classes.

And there is a broader digital-policy analogy here. Under the EU’s Digital Markets Act, Meta has interoperability obligations for WhatsApp Messenger, with BEREC noting that Meta has had to comply since 7 March 2024. The legal regime is different, but the principle is instructive. Access to an important platform should not require dependence on that platform’s native application environment. Interoperability matters because it limits lock-in and restores user control over the interface layer.

Fixed income should be heading in the same direction.

Venue-native workflow tools should be available, not unavoidable.

There is a serious counterargument, and it should be taken seriously.

Push de-perimeterisation too far and the market risks drifting into a world of quasi-venues, patchy controls and weaker surveillance. The tape does not solve every problem. It helps on transparency. It does not magically solve conduct, conflicts, access or market abuse.

So the answer is not to tear the perimeter down.

The answer is to draw it around the right things.

That means a layered model.

Keep a hard perimeter around genuinely multilateral trading, formal trade formation, venue operation, surveillance-critical functions and reporting obligations. Keep strict expectations around auditability, best execution accountability and governance.

But outside that hard perimeter, allow the buy side far more room to build and operate its own intelligent workflow layer: RFQ orchestration, protocol abstraction, AI-driven dealer selection, smart routing, negotiation support, pre-trade analytics and cross-channel execution logic.

In that model, the venue remains essential. It just stops being the default owner of the buy side’s execution brain.

The next phase of fixed income market structure should not be another crude attempt to stuff more behaviour into old entity labels.

It should be a rebalancing.

Let the tape carry more of the transparency burden.

Let the buy side own more of the workflow burden.

Let venues compete on liquidity access, execution quality and protocol design, not on how effectively they can make their UI and workflow stack unavoidable.

That would be better for innovation. Better for operating model resilience. Better for AI readiness. And, frankly, better aligned with how modern electronic markets ought to work.

The venue should be where I execute.

My smarts should stay in my stack.