Agentic artificial intelligence startup Kubiya Inc. today announced the imminent launch of the industry’s first “deterministic” composer as a way to ensure organizations can fully trust their AI agent deployments.

The startup said Kubiya Composer is different to existing probabilistic platforms for managing AI agents, because it offers more structured and predictable orchestration powered by the company’s proprietary Context Graph.

According to Kubiya, though AI agents are supposed to be all the rage these days, the reality is that many organizations are wary of leaning too heavily on them to automate critical tasks because they fear the lack of transparency around how they work. The challenge is that many existing AI agents make different decisions each time and struggle with “hallucinations,” which means they can’t be trusted with customer-facing applications.

Kubiya, which began life as a developer of generative AI assistants for DevOps teams, says Composer solves this through deterministic execution. That ensures that the same input always results in the same output, meaning AI agents become more predictable and trustworthy.

Kubiya co-founder and Chief Executive Amit Govrin told SiliconANGLE that his company’s deterministic approach goes far beyond existing AI agent orchestration platforms such as LangGraph and CrewAI. “These tools are great at the application layer, but they lack the depth, control and security needed at the infrastructure level to truly support enterprise-grade environments,” he said.

The lack of control explains why AI agents are still something of a rarity in enterprise deployments today, Govrin believes. Until organizations can trust that AI agents are making the right decisions, every time, they’re not going to fulfill their potential, he said.

“Specifically, they need verifiable trust that AI will do what it’s supposed to do, securely and reliably,” he said. “That’s exactly where we shine.

Govrin explained that the company’s deterministic composer provides engineering teams with an agentic orchestration layer that brings the rigor of traditional software development to AI workflows, meaning they can move away “black-box” agents that operate in much more mysterious ways. With its deterministic execution, the Kubiya Composer ensures that every action by an AI agent is testable, observable and repeatable, eliminating any surprises.

This is enabled by the continuously updated Context Graph, which provides what the company terms “organizational cognition.” It’s a real-time, vectorized representation of an organization’s systems, codebase, identities, workflows, knowledge assets and infrastructure state that provides full context into every decision an AI agent makes.

Security and transparency is enhanced through Kubiya Composer’s zero-trust architecture, which provides role-based access controls for AI agents with audit logs and policy enforcement to ensure full compliance.

Using the Kubiya Composer, organizations will be able to delegate some of the most complex information technology engineering tasks to AI agents. Govrin gave examples such as the creation and modification of infrastructure environments, the execution of deployment pipelines and interfacing with version control, ticketing, observability and continuous integration/continuous deployment systems.

“This feels like the moment the market’s been waiting for – a real answer to the question, how do we trust AI in production?” he said.

Kubiya said it’s inviting AI engineers, architects and other technical leaders to join the early access program for Kubiya Composer, available on Product Hunt from Friday, so they can start experimenting with more reliable, deterministic AI agents.

Image: SiliconANGLE/Dreamina