The Collapse of the Capital Markets Software vendor – and the rise of the Trading Desk

What Happens When Your Competitor Ships Features 100 Times Faster Than You?

Start with the number, because the number is the whole argument. Not 10% faster. Not double. Not a hundred times, but 1000x speed of delivery.

• Their head of trading asks for a workflow on Monday. It is in UAT on Tuesday.
• Their PM has an idea over coffee and is testing it that afternoon.
• Every gap in their platform closes the week someone notices it.

Yours sits in a vendor backlog behind an eighteen-month lead time and a change request form. Most of the industry is about to find out what that feels like. From the wrong side.

The desk starts asking again

The trading desk stopped asking for things years ago. Not for lack of ideas. Because asking cost two years and a backlog that never cleared. Every client we have carries a thousand changes they would make tomorrow if the effort of change was cheap. The whole market sits inside a terminal designed twenty years ago, and the benchmark for replacing it is parity with the old one.

That is not a market that stopped wanting. It is a market that stopped believing.

Collapse delivery from months to hours and the belief comes back on the spot. I watch it in every demo. Someone says “it would be nice if” and it exists before the next meeting. The second time, they stop saying “nice if” and start saying “I need.” By the third, they are redesigning their desk out loud.

That is not a software upgrade. That is people remembering what they are allowed to ask for.

Then everything downstream moves

The buying decision dies. Vendor selection has been a checklist for two decades: does it do what my current system does, is the vendor solvent. When features are immediate, the checklist is worthless. Exactly one question survives: does this platform let me build what I have not imagined yet?

Whole software categories die. TCA, surveillance dashboards, pre-trade analytics, reporting layers. If the data is already in the trading system and a screen over it takes an afternoon, nobody buys a separate product, runs a separate integration, pays a separate licence. Whoever holds the data is fine. Whoever sells a software layer on top of someone else’s data gets obliterated.

Roles flip. When code is no longer the constraint, the valuable person is whoever knows exactly how the workflow should behave. That is the trader. That is the QA analyst who has spent ten years knowing what correct looks like. Engineers become architects. Clients become product managers. We are retraining our QA function, not firing it.

And the vendor market breaks. Two hundred people can now out-deliver ten thousand. The vendors that own this market are debt-laden, run for EBITDA rather than product, and shedding the experienced people who could build anything new. They are trapped. Telling ten thousand staff to adopt a technology that makes most of their roles redundant is not a change programme. It is a mutiny. Their customers get slow progress and rising collapse risk. Their staff get to fear AI from the inside. Nobody wins except the firms that leave.

Qanvas

This is not a prediction. It happened to us this year. This is the first time we are saying so publicly.

In January the market decided AI would eat software vendors. Our peers explained at length why trading technology was too complex for that to apply to them. Nonsense. We took our web trading front end, two years of conventional development, and threw it in the bin.

The goal was never an AI-first system. No trader cares how their software was written. The goal was four things this industry has treated as mutually exclusive for twenty years. Zero defects at any speed. A front end that stays responsive no matter how much we change it. No compromise on user experience, ever, because a trader forgives a missing feature and never a slow blotter. And delivery two orders of magnitude faster. Days not months. Weeks not years.

Every process we had made those four contradictory. Speed cost quality, quality cost speed. So we started from a blank sheet, designed for all four at once, and the only way to get there was to take people out of the parts of the pipeline where people are the bottleneck.

The result is Qanvas. A trading front end built entirely from natural language, on the same deterministic core that has carried our clients’ order flow for twenty years. Six weeks to rebuild what took two years. Five months in, a team of effectively 1.4 people has shipped around 950 features. Fifty a week per head. Across the people we already employ, that is a thousand features a week or more. In the past we advertised 3,500 features a year in our sales pitch. Our problem now is that we build faster than we can think of what to build, 3,500 features seems laughable, a week of work at most.

It is a “canvas”. Everything on it can be redrawn, and increasingly the client holds the brush. It is in production. You can buy it today.

Why it works, in one sentence: Quod was built around a single normalised message bus because we believed trading would become automated, and a build layer needs exactly what an automation layer needs. Everything the platform can do, through one interface. We are not an AI company. We are a software company that finally has the delivery pipeline it always needed.

Three ways in

The front end is a tactical lead of six months to two years. Every vendor will eventually have an AI-generated interface and they will all look the same. What will not converge is underneath: twenty years of know-how about how venues, regulators, edge cases and clients actually behave, expressed as software that always works. The knowhow and software pipeline that we have curated over 10+ years cannot be prompted into existence. There is no shortcut.

Qanvas as product. Tell us what your desk needs. We build it, we product-manage it, it lands in your UAT that week.

Qanvas as framework. Use our build process yourself. Define it, prototype it, iterate it, hand us something that drops straight into our pipeline for production hardening. No developers required.

Quod as core. Take our APIs, build your own front end with your own AI, never talk to us about interfaces again. You get order lifecycle, risk, audit, connectivity and determinism. You build the workflow. Nobody bypasses the controls, because everything still routes through the trading software.

What breaks

Everything, and we are still rebuilding it.

We do not know how a head of trading runs a floor where every trader can reshape their own screen. Today the desk owns the requirement and a locked default workspace is always there. That is a guess.

We do not know whether the end state is one product or one Qanvas per client. Everything merges to a single trunk behind feature flags. Whether that holds once clients build in volume, and what it costs in agents and tokens to maintain divergent branches forever, is open.

We do not know how far this goes into the back end. Front end and edges first, because they are isolated and stateless. Order lifecycle, positions and risk are harder. I am not promising a date.

We are lifting off human-defined processes and we do not know where we land. Anyone who tells you they do is selling something.

Stand still and get crushed

I am afraid for this business in three years. I should be. The barrier to entry is falling and the only thing protecting any vendor is trust earned over decades and know-how that cannot be prompted into existence.

Fear is a poor reason to stand still. You either wait for the change and get crushed by it, or you sprint as hard as you can to be the change. I am not going to be put out of business by the AI future. We decided to “be it” instead.

The industry has spent two years looking for AI in the chatbot. It was never there. It was never going to be there. It is in the pipeline that turns a sentence into working, tested, responsive software in days. Once that exists, every workflow, every role and every assumption about what a trading system can be is up for renegotiation.

Tell us what you need. We will deliver it to you by next week.

Back to top