The AI architecture question every enterprise is about to have to answer

For three years, the AI industry has been selling enterprise buyers on the same four things. Bigger. Faster. Higher scores. Longer memory. And enterprise buyers, under pressure from their boards to do something about AI, have largely been buying it.The problem is that those factors may not fully capture the outcomes it considers most relevant. Or whether the system keeps working when the world stops cooperating.Vertusis an AI company based on the Isle of Man, co-founded by Julius Franck, Alex Foster, and Michal Prywata. Its founding team brings together quantitative systems architecture, algorithmic trading infrastructure, regulatory frameworks, and cross-domain engineering across medical robotics, space systems, and biotechnology. According to the company, its system has been deployed in live financial market conditions, with the company reporting an annual return of just over 50 percent and a Sharpe ratio above 2.0, in conditions that included the largest two-day market loss in market history.The numbers provide one point of reference. What matters more for enterprise AI decision-makers is why.Michal Prywata, who leads technology architecture at Vertus, has explained the differences clearly. Almost every AI system currently being deployed in enterprise environments is built on what he calls conditioned response at scale. The AI system trains on historical data, identifies patterns, and generates outputs based on statistical continuation. In stable environments, this produces results that are genuinely useful. In environments where the structure of the problem changes, the system continues merrily along on its way and generates statistically probable outputs that are wholly based on conditions that no longer apply. Fluently. Coherently. Incorrectly.The enterprise failure mode is the same as the financial one. A healthcare system that generates confident-sounding treatment recommendations based on patient profiles that match historical patterns, but without recognizing that the patient’s current condition doesn’t fit the mold anymore, is now in a very real life-and-death situation where its advice is putting real people in real danger. A supply chain system that optimizes routes based on supplier relationships that no longer exist as of last quarter is going to create more than a traffic jam. A legal AI that produces precedent analysis based on regulatory frameworks that were amended after its training cutoff was loaded into the system and has now built a house of cards that’s doomed to collapse. In each case, the system isn’t broken. It’s doing exactly what it was designed to do. Designing assumptions that stopped applying because the world changed. And that’s the problem.Prywata’s background is relevant here. The Vertus intelligence reflects those design principles. Not a fixed model trained once and deployed. But a system that, according to the company, is designed to adapt its reasoning to different problems, incorporate prior information, and adjust when earlier assumptions no longer appear to apply.The platform model Vertus has built may also warrant further consideration from an enterprise deployment point of view. Vertus doesn’t operate as a fund. It functions as the cognitive and execution backbone for partner funds that implement its cognitive architecture. This intelligence is available simultaneously across multiple deployments rather than being consumed by only one user. And in this way, each deployment stress-tests the intelligence under different conditions, which the company describes as a way of continuously updating its understanding rather than relying on fixed models.This is a different theory of how enterprise AI infrastructure should work. Not proprietary models locked inside single deployments and depreciating as conditions drift away from their training data. But a shared cognitive infrastructure that appreciates and culminates knowledge through use. Each deployment stress-tests the intelligence under different conditions, continuously expanding its real-world experience rather than working from a fixed and increasingly outdated position.The Vertus API is now live for developers and institutions beyond finance. Healthcare. Scientific research. Supply chain management. Infrastructure. The same system that, according to the company, was associated with a reported Sharpe ratio above 2.0 in the most structurally challenging market year in recent memory, is being extended into every enterprise field where the structure of the problem changes while the system is working in real time.The enterprise AI question that every CTO and systems architect is going to face in the next eighteen months isn’t which model has the highest benchmark score. Its intelligence is able to maintain coherent and responsible reasoning when the situation changes and the rug is pulled out from underneath it.According to the company, Vertus addressed that question in a live market environment, which it characterizes as a rigorous test.The enterprise equivalent of that test isn’t coming in some distant future. For most organizations using AI in mission-critical systems, it’s already here. The only question left is whether the intelligence you chose is going to be able to pass it.The information provided in this article is for general informational and educational purposes only. It is not intended as legal, financial, medical, or professional advice. Readers should not rely solely on the content of this article and are encouraged to seek professional advice tailored to their specific circumstances. We disclaim any liability for any loss or damage arising directly or indirectly from the use of, or reliance on, the information presented.VentureBeat newsroom and editorial staff were not involved in the creation of this content.

Raccontata daventurebeat.com