In many large companies, data science teams face significant delays in deploying advanced models due to bureaucratic processes. For instance, a data science team may spend six months developing a model that predicts customer churn with 90% accuracy, only to see it become inactive while awaiting approval from a risk review committee. This situation highlights a broader issue in enterprises where innovation in artificial intelligence (AI) is outpacing organizational processes.
The rapid pace of advancements in AI technologies, including new model families and evolving toolchains, contrasts sharply with the slower operational frameworks of enterprises. This has created a “velocity gap” where enterprises struggle to keep up, leading to inefficiencies such as missed productivity and compliance challenges.
Two primary trends contribute to this gap. First, the business sector has become the largest contributor to notable AI innovations, according to the Stanford 2024 AI Index Report, with increasing demands for compute resources. Second, a growing number of enterprises are deploying AI solutions, but governance measures have yet to be fully established, necessitating retroactive control measures.
Compliance regulations, such as the EU AI Act, further complicate matters. Enterprises face upcoming deadlines for various governance obligations, yet many lack the necessary frameworks to manage them.
The main challenge for enterprises lies not in model development itself but in complying with audit requirements. Current governance policies, often designed for static software, inadequately address the complexities of AI models. This results in increased review times and complications in proving model adherence to guidelines.
Companies aiming to bridge the velocity gap should implement practical frameworks that prioritize governance, including establishing standardized processes for risk assessment tied to use-case criticality. A structured approach to governance can foster a balance between rapid innovation and necessary compliance, allowing companies to maintain productivity without excessive delays.
Source: https://venturebeat.com/ai/heres-whats-slowing-down-your-ai-strategy-and-how-to-fix-it