The desk should own its workflow.

That sounds obvious, yet much of the market still behaves as if workflow is something to be inherited from venues, brokers or legacy vendor design. It is not. The buy side’s core execution logic should sit in the firm’s own stack, where it can be governed, adapted and improved over time.

If the most important parts of decision-making live somewhere else, the desk does not really own the workflow. It merely uses it.

A multi-asset stack should not feel like three separate products held together with optimism.

Fixed income, FX and equities each have different microstructure, liquidity dynamics and execution protocols. That is fine. What is not fine is forcing traders and firms to operate through completely different control models, user experiences and support structures for each one.

A modern stack should provide a coherent operating model across asset classes while still respecting the specifics of each market.

The stack should coordinate how orders and RFQs are staged, sequenced, routed, monitored and adjusted across channels.

That is not back-office plumbing. It is the execution brain.

In a modern trading environment, orchestration determines how the desk interacts with liquidity, how it controls information leakage, how it handles exceptions, and how it decides when to remain bilateral, when to widen distribution and when to automate.

A stack without orchestration is just a collection of doors.

Venues, dealers, algos, brokers and data services should be consumed through robust APIs wherever possible.

If access to liquidity depends on living inside a proprietary screen or workflow environment, that is a design weakness, not a competitive advantage. Screens still have their place. But they should be optional front ends, not the place where the desk’s operating model goes to live and die.

A modern stack should treat external providers as interoperable endpoints, not separate little kingdoms with their own laws and customs.

The trader should see the right context at the right moment.

Not more data. Better-timed data.

Contextual intelligence means surfacing the relevant information in the flow of execution: liquidity conditions, protocol suitability, dealer behaviour, venue characteristics, event risk, historical outcomes, market regime and order urgency. The goal is not to bury the desk under more widgets. It is to make the workflow smarter and more discriminating.

A good stack does not just show the market. It helps the desk interpret it.

Rules, approvals, thresholds, exception handling and audit should be governed centrally rather than duplicated across tools, venues and workarounds.

This matters for obvious reasons such as compliance and auditability. It also matters for much more mundane reasons, which is often where the real pain sits: supportability, consistency, change control and the ability to sleep at night.

A fragmented control model is how firms end up with contradictory rules, inconsistent behaviour and a dependency on tribal knowledge. That is not resilience. It is institutionalised wishful thinking.

The stack should help the desk control when and how intent is externalised.

This matters across all asset classes, but especially in markets where liquidity is episodic, price formation is less centralised and signalling risk is material. The stack should support controlled escalation, intelligent protocol choice, sequenced outreach and selective exposure rather than treating every order as a candidate for immediate broad dissemination.

The ability to manage information leakage is not a side feature. In many cases, it is a core part of execution quality.

Execution data should feed back into decision-making.

That means trades, quotes, dealer responses, venue outcomes, timings, exceptions and market conditions should be captured in a way that improves future routing, protocol choice, dealer selection, model tuning and post-trade assessment.

A stack that does not learn is just an expensive front end with good branding.

The point of a modern execution stack is not only to support today’s trade. It is to make tomorrow’s trade better.

AI should sit in the workflow itself.

Not bolted on later as a cheerful sidecar. Not sprinkled on top of a broken operating model like parsley on burnt fish.

AI should help the desk prioritise, summarise, recommend, simulate and automate within the actual flow of execution. It should support decisions such as who to ask, how wide to go, when to switch protocol, how to interpret a response pattern, or whether current conditions resemble a known scenario.

That only works if the stack was designed to support contextual intelligence in the first place.

A better stack is not the one with the most components. It is the one that reduces duplication, avoids unnecessary dependencies and makes change safer.

Too many firms still confuse sophistication with sprawl. A modern stack should simplify where it can, centralise where it should, and externalise only where it makes strategic sense.

The buy side does not need more moving parts for the sake of looking modern. It needs an operating model that is robust under pressure, governable at scale and adaptable without a six-month support crisis every time something changes.

Dealer selection, wave logic, timing and fallback paths should be controlled in the desk’s own environment.

That includes decisions about whether to stay bilateral, how to sequence outreach, when to widen the panel, how to compare responses and how to adapt in real time. RFQ is not just a protocol. It is a workflow, and the desk should own it.

Fixed income is not one channel pretending to be a market.

A good stack should support bilateral, RFQ venue, voice-assisted and algorithmic workflows without forcing one channel to dominate the entire operating model. The point is not to declare a winner. It is to give the desk flexible control across multiple forms of liquidity interaction.

Showing your hand too early is often the whole problem.

A fixed income stack should actively help the desk minimise unnecessary information leakage, manage distribution carefully and sequence interaction with intent. This is where workflow design matters more than slogans about digitisation.

The desk should be able to combine multiple forms of information in one decision layer: axes, historical dealer behaviour, quote quality, response patterns, venue characteristics, market regime and current conditions.

If the trader has to stitch that together mentally across four screens and two chats, the stack has failed.

Execution choices in fixed income are highly context dependent.

Urgency, market tone, issue characteristics, liquidity regime, volatility and client sensitivity all affect the right path. A good stack should adjust workflow based on scenario, not just instrument type. That is where execution starts to become genuinely intelligent rather than merely electronic.

The system should know when to use streaming liquidity, when to request quotes and when to split or stage flow across approaches.

FX may be more electronically mature than fixed income, but that does not make protocol choice trivial. Good workflow means the stack supports that choice intelligently.

Not all liquidity is equal.

The desk should be able to assess provider quality through reject rates, fill behaviour, slippage, skew behaviour, latency characteristics and other measures that actually affect trading outcomes. Raw availability is not the same as execution quality.

FX is global, continuous and event-sensitive.

Routing, automation and decision support should adapt to time zone, market session, known event risk and current liquidity conditions. A stack that treats London lunch and Asian open as equivalent states of nature is asleep at the wheel.

A smart stack should help firms identify where flows can be offset, crossed or managed more efficiently before externalising risk.

That is not just an execution benefit. It is also a balance-sheet, cost and control benefit.

FX rarely sits alone.

It often interacts directly with fixed income exposures, equity flows, macro-event risk and portfolio rebalancing activity. A modern stack should reflect those linkages rather than treating FX as a silo with its own small republic of workflow.

Broker and algo selection should be measurable, explainable and adaptable.

Too many desks still operate as if some algo choices are handed down by folklore. A modern stack should evaluate broker and algo behaviour continuously and feed that intelligence back into allocation and routing decisions.

Smart order routing should be governable and visible.

No one needs yet another black box wrapped in vendor confidence. The desk should be able to understand how routing decisions are made, how they can be adjusted, and whether they are delivering what they claim.

TCA should inform decisions while trading, not arrive afterwards as a tidy PDF obituary.

Equities already has the richest post-trade analytics tradition of the three asset classes. The next step is to make those insights more live, more contextual and more decision-relevant.

The stack should account for auctions, earnings, rebalances, index events and corporate actions in real time.

The goal is not just to know that events exist. It is to make them part of the execution decision layer.

The same control framework should span DMA, algo, program and manual workflows.

The desk should not have one logic for high-touch trading, another for low-touch flow and a third for everything people hope compliance never notices. Control, governance and intelligence should travel across execution styles.

The answer is not another front end.

It is not another point solution pretending to be strategic. It is not another venue screen with a fancier settings menu. It is not another layer of workflow duplication wearing the badge of innovation.

The way forward is to treat the buy-side trading stack as an execution fabric built around a few clear principles.

First, centralise the intelligence. Workflow logic, orchestration, controls, contextual analytics and AI should sit in the firm’s own environment.

Second, abstract the endpoints. Venues, dealers, algos and brokers should be integrated as execution endpoints, not separate workflow worlds that each demand their own little empire of settings, support and exception handling.

Third, normalise the workflow. The desk should operate through a consistent decision layer across asset classes, even where market-specific execution logic differs underneath.

Fourth, build feedback into the system. Every trade, quote, response, exception and market state should improve the next decision.

Fifth, design for augmentation first, automation second. The goal is not to remove the trader. It is to make the trader and the desk materially more effective, with automation applied where it earns the right to exist.

That is what a modern execution fabric looks like.

Not one more screen.

Not one more silo.

Not one more vendor telling you that this time the fragmentation will feel strategic.

The future multi-asset desk will not be defined by how many venues, algos or data feeds it can technically access. That has become table stakes.

It will be defined by whether its intelligence sits in one coherent stack or leaks out into everyone else’s.

That is the real dividing line.

A modern buy-side trading stack should not be a collection of screens. It should be an execution fabric: interoperable, scenario-aware, data-driven and AI-ready by design.

Venues, brokers and dealers will remain essential. Of course they will.

But they should be destinations for execution, not the place where the buy side’s workflow intelligence goes to live.

The buy side should demand more than multi-asset in name.

It should demand a stack that lets it think, decide and improve in one place.

Its own.