AI Sovereignty – A Strategic Priority for Capital Markets
Thought Leadership
Authored by
Audience
Read Time
AI Sovereignty is no longer just a concern for governments. Any large organisation with proprietary data and workflows needs to consider how they can go all-in with AI, whilst protecting their IP and managing their cost base.
As Frontier Labs (Anthropic, Open AI) continue to enter the vertical sector application layer, any organisation with IP to protect needs to think twice about their reliance on models coming out of these labs.
Case in point: when Claude Design was released, Figma had every reason to cry foul. Not only were Figma’s clients feeding intelligence to Claude daily with their pre/post training and inference, but the two companies worked closely together until that point. Three days before Claude Design’s release, Mike Krieger – an Anthropic executive– sat on Figma’s board.
Capital Markets organisations place tremendous value on the IP embedded in their data, models and business processes. As the spread between Open-Source and Frontier models continues to shrink, the AI strategy that might have worked six months ago needs to be reconsidered.
Enter the Engine AI Enterprise Harness.
How the Engine AI Enterprise Harness Works
A core part of our Intelligence Engine is the Enterprise Harness, that has two components:
Data: The data harness connects all data relevant to any workflow into our Intelligence Engine. Organisations can finally deliver intelligence building on decades of investment in data and digital transformation, without expensive and lengthy data migration projects.
Analytics: On top of the connected data, the harness applies AI to the Machine Learning (ML) models developed over decades. The harness routes workloads efficiently between trusted ML and Open-Source models for the majority of activity, and Frontier models for the minority of tasks that require cutting edge capabilities.
80/80/20
More than 80% of workloads do not require AI inference.
AI can be used for three main tasks, considered by the industry as AI-generation: data retrieval; data analysis; and triggering a new action. The vast majority of Capital Markets activity is repetitive and does not need AI inference, with the exception of novel questions or analysis - these constitute less than 20% of activity.
Our proprietary Vibe Analytics allow users to build their own workflows with AI, leveraging templates and other capabilities that address frequently asked questions and tasks that cover the majority of activity we see for every Capital Markets persona.
More than 80% of AI workloads do not require a Frontier model.
Of the 20% of workflows that require AI inference, more than 80% do not require the latest, most expensive model. Open-Source models, fine-tuned for each Enterprise, can carry the bulk of the load.
As a result, in most cases, Frontier models are used in less than 10% of workloads. In addition, when Frontier models are used for structured data analysis, we do not send proprietary Enterprise data to LLMs, providing further protection and efficiency.
The Upshot
Protect your IP
Save >90% off your LLM bill
Deliver more reliable, faster results to users
Minimise supplier dependency
Minimise geopolitical risk.
To see what this looks like for your firm, talk to the Engine AI team.
