My business, which relies heavily on AI to help modernize and document legacy application code, has settled on an approach focused on managing the AI solutions that users are employing and how they are using them. Instead of allowing employees to use their chosen AI model without guardrail type, they route all calls to the AI model on the network through the management platform we have built. The platform mediates calls and responses to flag people who may present security or compliance risks.
This approach allows the platform to support offerings from all major vendors (Openai, Anthropic, Google, etc.), giving users the flexibility to use a variety of AI models. It also allows an open-ended approach to how employees use AI. They are not limited to a limited range of defined tasks and solutions. They can issue the prompts they want on a completely self-service basis. But at the same time, using routed calls over the platform allows employees to enforce enterprise-level control of what they do with AI tools.
This is what we hope for the future of AI-powered self-service to look like an IT department. Rather than placing third-party AI tools in the hands of users and leaving them to their own devices, businesses need to implement hubs that manage how employees interact with AI. They will want to keep AI use cases flexible while maintaining their ability to mediate security, compliance and performance risks.
