Consolidated tape is attracting attention because it is new, important and politically visible. That makes sense. Firms are looking at how it fits into their architecture and operating model, and many will be tempted to take the easy route: wire it into the EMS, expose some of it in the OMS, perhaps make a cut-down version available elsewhere, and move on.

That will be the easy answer.

It will rarely be the best one.

Because consolidated tape is not just another market data widget. Nor are pricing feeds, quote feeds, axes or signal feeds. These are not valuable merely because they can be viewed. They are valuable because they can be shaped, compared, combined, contextualised and turned into something more useful than raw display.

That is where too many firms still undershoot.

The integration model remains application-centric when it should increasingly be intelligence-centric.

The mistake is simple.

A strategic data source arrives. The firm asks which application should consume it first. The design gets led by the screen rather than by the use case. The data is shaped for display rather than reuse. The output is human-facing rather than workflow-facing. The result is another grid, another tab, another partial view, and another small increment in application sprawl.

The trader gets more data.

The desk does not necessarily get more intelligence.

That distinction matters.

Traders do not need another static grid. They need data in a form that helps them prioritise faster, sequence actions more intelligently, reduce information leakage and improve outcomes.

The old question is: what can we show?

The better question is: what can this data help the desk do?

This is not an argument against OMS or EMS. Both are essential parts of the operating model and both should consume relevant outputs.

But neither should usually be the strategic home of the data itself.

The OMS is naturally rich in order context, lifecycle state, investment constraints, approvals and workflow initiation. That makes it important. It does not make it the right place to trap strategic market data that needs to support multiple downstream workflows, analytics, signals and models.

The EMS is naturally where traders interact, assess execution options and control workflow at the point of execution. That makes it the obvious landing spot for new feeds. Again, understandable. Again, not enough.

If strategic data is embedded primarily in the OMS or EMS, it is too often shaped for immediate visual consumption rather than firm-wide reuse. The result is local convenience at the expense of wider value.

OMS and EMS should consume intelligence.

They should not usually be where that intelligence is created and trapped.

The better pattern is straightforward.

Strategic market data should land first in a shared intelligence and orchestration layer, not directly in a screen-led application silo.

That layer should do far more than basic ingestion. It should provide the harness that turns raw data into usable intelligence across workflows, decisions and models. That means canonical modelling, normalisation, enrichment, event processing, entity mapping, time alignment, feature extraction, scenario detection, signal generation, memory and state handling, and model-ready outputs.

Only then should results be delivered into the OMS, EMS, dashboards, analytics tools, control frameworks or AI workflows that actually need them.

That is the difference between integrating a feed and building a capability.

Once data is shaped centrally, it can support trader-facing context, machine-facing features, pre-trade signals, routing logic, dealer and venue benchmarking, post-trade assessment, surveillance, scenario-aware recommendations and AI-driven prioritisation.

That is a very different end state from saying, “we added it to the EMS”.

This is where the old integration pattern becomes actively unhelpful.

In the old model, it was at least arguable that getting a feed into the OMS or EMS was enough because the trader could do the synthesis manually. The human was the harness. The human carried the context, remembered the history, handled the edge cases and chose the action.

That argument weakens quickly in an AI-enabled target state.

Because the challenge is no longer just exposing data. It is building the operating layer around it.

For enterprise AI, the real problem is rarely model access on its own. It is whether the firm has the right harness around the model with the right tools, the right context, the right memory, the right retrieval, the right workflow integration and the right controls.

That is where many firms still fall short.

A strategic market data source is not valuable to AI because it can be displayed in another grid. It becomes valuable when it can be combined with order context, execution history, entity relationships, venue behaviour, dealer responses, constraints and outcomes in a way that is structured, reusable and trustworthy.

Long context alone does not solve this. Dumping more tokens into a prompt is not memory. Nor is it a serious enterprise architecture. Retrieval quality still matters, especially where dates, numbers, sequencing and lineage are involved. Tool use still matters when actions need to be grounded in real workflows rather than chat-shaped improvisation.

The firm’s edge will not come simply from choosing a strong model.

It will come from how well it wires data, context, tooling, memory and workflow together around that model.

That is the real design problem.

Once firms accept that strategic market data should be shaped in a shared intelligence layer rather than simply dropped into an application, the next question is how to implement that target state.

Build makes sense where the firm wants real control over canonical models, enrichment logic, feature engineering, signal extraction, orchestration and AI use cases. The upside is flexibility, portability and differentiation. The downside is that building well is hard, and many firms like the sound of strategic control more than the reality of funding it properly.

Buy makes sense where speed, lower implementation burden and faster visible consumption matter most. That can be sensible, particularly for narrower use cases. The risk is that convenience becomes architecture, and strategic context disappears behind someone else’s product assumptions.

For many serious firms, hybrid will be the most sensible answer. Use vendors for genuinely commodity functions such as feed handling, onboarding, entitlement, distribution and basic normalisation. Retain ownership of the parts that create edge such as canonical models, cross-source reconciliation, feature and signal layers, orchestration logic, workflow integration, feedback loops and AI use cases.

Hybrid is not indecision. Done properly, it is disciplined separation between what is commodity and what is strategic.

Before deciding, firms should ask a few blunt questions.

Those questions tend to reveal very quickly whether the target state is real or whether the firm is just putting another screen-shaped plaster over a broader architecture problem.

The clearest lesson is not that the industry lacks access to data.

It is that the industry still too often lacks the right mental model for what strategic market data is for.

If the feed is treated as an application feature, the result is more display.

If it is treated as a firm-level intelligence asset, the result can be better workflow, better decisions, better controls and a stronger foundation for AI.

That is the real architecture choice.

Not OMS versus EMS.

Not even model versus model.

Display versus intelligence.

The firms that get the next phase right will not be the ones that merely integrated more feeds into their OMS or EMS. They will be the ones that built a stronger harness around data, context, tooling, memory and workflow, and turned that into a shared intelligence layer for decisioning and AI.

Consolidated tape is simply the latest opportunity to choose correctly.

The same lesson applies to axes, pricing, quote and signal feeds too.

The question is no longer where to display the data.

It is how to shape it, combine it and operationalise it so it drives better actions and better outcomes.

That is where the real value is.