By Rohan Douglas, CEO of Quantifi
As technology, and AI in particular, continues to evolve and reshape the finance industry, trading desks are reconsidering how they source, maintain and extend their risk, pricing and analytics model libraries.
Historically, this has often been framed as a build-versus-buy decision. Do you develop and maintain your own pricing and risk models, tuned to how your firm actually trades, and take on the cost of the developers and quantitative specialists required to support them? Or do you adopt a validated library from a vendor at a fraction of the cost, but with less flexibility?
In practice, the choice is rarely that simple. Many firms combine vendor technology, proprietary models and internally developed tools. AI has not fundamentally changed the underlying economics of that decision. Building and maintaining core enterprise analytics remains a specialist, resource-intensive undertaking. What AI is changing is who can build on top of that foundation.
AI-Enabled Business Users
AI coding assistants have changed what it takes to create analytical tools and, in doing so, have lowered the barrier between a business idea and a working application. For the first time, what can be created is no longer limited by who can code. Imagine an AI or large language model that understands the breadth and depth of a model library, including its coding requirements, conventions and quirks.
Now imagine that capability in the hands of a trader or portfolio manager. A concept or idea could become a working tool, workflow or dashboard in an afternoon. Previously, this process might have taken months, required specialist developers and depended on the business agreeing that it was a development priority.
This does not mean that business users can suddenly replace the model library underneath the tool. The pricing, risk and analytics still need to come from a trusted, validated and well-maintained foundation. The bigger change is in business-user enablement. AI can make those capabilities easier to access, configure and apply without requiring a developer to translate every request.
It is true that this may initially be seen as an efficiency for a small number of traders. However, there is a realistic potential for AI to move competitive advantage away from firms with the deepest engineering resources and towards those that can enable their business users to test, visualise and validate ideas quickly.
Automated Trust Over Manpower
There are serious considerations for the finance industry as AI tools become available for front-office teams to create applications and workflows for themselves. Perhaps the most important is ensuring that probabilistic AI tools are grounded in deterministic, validated and auditable analytics. A major mistake firms may make is placing too much trust in AI agents that are not properly constrained or connected to a reliable analytical foundation.
The rise of AI coding assistants and terms such as “vibe coding” have been likened to a modern magic that allows users to build whatever they can imagine. However, especially for trading desks dealing with large sums of money, these tools must be handled with caution. AI is very good at producing something that looks right, but whether it is right depends entirely on what sits underneath it.
A convincing pricing tool built on poor analytics is simply a convincing way to misprice a position. In trading, the cost of a wrong number is not a bug report. It is a loss. A general-purpose AI model may invent numbers, apply the wrong pricing methodology or produce different answers when asked the same question more than once. Being able to generate repeatable and auditable figures from a model library that can be trusted remains critical.
The objections we hear most often from the industry are grounded in security, intellectual property and a broader uncertainty around how far these tools can be trusted. These are the right questions to ask. Building a tool yourself does not remove accountability for the output it produces. For trading desks, ensuring the right guardrails are in place will ultimately be a greater competitive differentiator than the speed at which tools are rolled out.
Buy… And Build?
The question of Build vs Buy is not disappearing but is distilling into a sharper question. Instead of one or the other, it requires both the acquisition of a strong foundation which you can build analytic tools on top of. Firms that grasp this concept, equipping their people with a foundation worth building on, putting the right guardrails in place, and rethinking how they pay for it all, will be the ones with a competitive advantage.
Subscribe to our newsletter
