Enterprises are banking on cutting-edge AI models to transform their business but how do they achieve this without compromising on security or resilience?
Anthropic’s Claude has particularly captured the imagination of enterprises, with Menlo Ventures estimating that it accounts for 40% of enterprise large language model (LLM) spend.
That rapid growth may have been fuelled, in part, by developers buying subscriptions outside of regular procurement processes.
“It was never that the model itself was risky,” Artur Schneider, CTO, AWS Business, at German transformation specialist Skaylink, said in a recent podcast series, Journey to AI.
“But,” he continued, “the access was happening in a blind spot outside normal governance. That so-called shadow AI environment is exactly why access has become a governance question in its own right.”
This tension is most obvious in regulated industries, financial services and health, adds AI GTM leader Daniel Harrison at AWS.
So, technology leaders need to consider the “distinction between data [that] stays within my AWS boundary or within my account versus data going directly to Anthropic-operated endpoints.”
From there, he says, “When you think about consuming cloud or consuming Anthropic models via AWS, there’s a plethora of ways that you can do that.”
Likewise, companies need to consider the question of sovereignty. Once companies map their data flows, they can consider the best model to ensure security and resilience, whether that’s confining data to a particular AWS region or accessing the AWS European Sovereign Cloud.
This is particularly key when it comes to agentic AI. Without that overall visibility and sound governance, multiple teams can end up building independent stacks, resulting in sprawl, which can present compliance issues.
Ultimately, says Harrison, success will come to those who can “master the mundane.”
“You need to build a strong foundation…that’s going to allow for you to really achieve some of the long-term results that you want.”
Explore how to build an AI platform that enables businesses to really understand their AI usage and deployments with Skaylink on AWS by listening to the full episode.
