Copyright SiliconANGLE News
Artificial intelligence, routed via an agentic AI platform, offers fewer miracles and more guarantees — but only where governance outruns hype. Organizations consistently agree that production-grade agents need a policy-aware control plane to manage access and risk. Whatever the stack, it should integrate control, context, lifecycle management and cost-aware observability to maintain auditability at scale, according to Raju Datla (pictured, left), chief executive officer of Fabrix.ai Inc. That is the perspective that informs Fabrix.ai’s own agentic AI platform. “It provides both control plane, [the] context plane and it also ensures that there is a governance and visibility into everything that the agents are doing,” Datla said. “It also provides built-in observability that ensures that you can monitor what the agents are doing, where they’re succeeding, where they’re failing, how much are they costing, so that you don’t have an AI cost model. Suddenly, [it is a] pay-as-you-go model; suddenly it warrants the cost.” Datla and Rached Blili (right), distinguished engineer at Fabrix.ai, spoke with Bob Laliberte, principal analyst at theCUBE Research, and Zeus Kerravala, principal analyst at ZK Research, at the “Agentic AI Unleashed: The Future of Digital & IT Operations” event, during an exclusive broadcast on theCUBE, SiliconANGLE Media’s livestreaming studio. They discussed Fabrix.ai’s own agentic AI platform seeks to close the AI value gap and operationalize autonomous operations. (* Disclosure below.) How an agentic AI platform can close the value gap Enterprises are adopting a common operating model that Fabrix.ai’s platform aims to directly support. Teams define personas that specify model access, approved tools and task-specific prompt templates; publish them in catalogs for reuse; and execute them non-interactively through agentic workflows. For this, interoperability is key — specifically an AI-programmable layer that lets humans or AI adjust easily within agentic information technology systems, according to Datla. “The data fabric [is] the AI programmable fabric where you can dynamically create new data, add new data sources, ingest new data, enrich data or reroute the data,” he explained. “That data can be in our platform, or you can pull the data from any IT systems, networks [or] infrastructure elements dynamically — and the data fabrics can be configured by humans or even by AI.” But teams don’t just spin up these agents — they curate reusable instructions that encode how specific tasks should be done. That’s where prompt templates act like operational runbooks, giving each persona a consistent, auditable way to execute, according to Blili. “You can think of a prompt template as a bookshelf that contains instruction binders,” he explained. “The agent … looks on its bookshelf and asks, ‘Do I have a manual that teaches me how to handle this particular kind of request specifically?’” In practice, Fabrix.ai promotes cost and usage to first-class telemetry. Owners can break down tokens and dollars by model, agent and job, trace tool calls end-to-end and correlate those runs with user feedback and evaluator scores, according to Blili. That stance treats AI workloads like any other production system — measured, audited and improved. “We treat AI as a first-class citizen at Fabrix.ai,” he said. “We’re basically able to instrument and keep track of pretty much everything that’s going on from a cost perspective and a usage perspective.” Here’s the complete video interview, part of SiliconANGLE’s and theCUBE’s coverage of the “Agentic AI Unleashed: The Future of Digital & IT Operations” event: (* Disclosure: TheCUBE is a paid media partner for the “Agentic AI Unleashed: The Future of Digital & IT Operations” event. Neither Fabrix.ai, the sponsor of theCUBE’s event coverage, nor other sponsors have editorial control over content on theCUBE or SiliconANGLE.) Photo: SiliconANGLE
