As enterprises grapple with managing increasingly complex hybrid infrastructure across multiple clouds, on-premises data centers, and edge locations, the promise of AI-driven automation has collided with a critical challenge: how to let AI agents reason and plan autonomously without compromising operational safety and control. Itential, a leader in infrastructure and network orchestration, is addressing this gap with FlowAIa new agentic layer that separates intelligent reasoning from deterministic execution, creating what the company calls a “closed AI-to-Action loop.”
In this exclusive VMblog Q&A, Peter Sprygada, Chief Architect at Itential, explains how FlowAI enables AI agents to understand intent, make decisions, and coordinate complex multi-step tasks while ensuring every action passes through a governed execution layer that enforces schemas, permissions, and full audit trails. Unlike AI-native workflow platforms that blend reasoning and execution, FlowAI’s architecture prevents hallucinated commands, privilege overreach, and unapproved changes from reaching production environmentsmaking it possible to safely automate high-stakes network and cloud operations. As FlowAI enters Private Customer Preview, Sprygada shares insights into the technical innovations driving this approach and the organizational challenges enterprises must navigate to embrace AI-supported infrastructure orchestration.
++
VMblog: As we begin, can you explain to our readers what Itential does and what market problem it solves?
Peter Sprygada: Itential provides an infrastructure and network orchestration platform for hybrid environments. The core problem we solve is operational complexity. Enterprises today manage networks and infrastructure across multiple clouds, on-premises data centers, and edge locations. Each has different APIs, tools, and workflows. Itential unifies these environments into a single orchestration layer, allowing teams to automate configuration management, compliance checks, and service delivery without building custom integrations for every system. We handle the translation between intent and execution across heterogeneous infrastructure.
VMblog: Itential is announcing FlowAI. What specific problem is Itential trying to solve when it decided to add an “agentic layer” to the Itential Platform, and why couldn’t traditional automation workflows handle it?
Sprygada: Traditional automation workflows assume a human provides structured inputs. They are not built to interpret open-ended intent, reason through ambiguity, or plan multi step tasks the way AI agents do.
The problem enterprises faced was clear: AI agents can generate intent and plans, but they cannot safely interact with infrastructure on their own. There was no controlled way to translate agent reasoning into real operational action.
FlowAI adds an agentic layer to the Itentil Platform that can understand intent, make decisions, and coordinate tasks while still relying on a governed execution layer for all actions. This allows enterprises to benefit from autonomous reasoning without compromising safety or control.
VMblog: With FlowAI, Itential seems to emphasize the combination of “agentic reasoning” and “deterministic execution”. Can you walk us through a concrete example of how these two concepts work together in a real infrastructure operation?
Sprygada: Imagine an enterprise preparing for a major network update across multiple clouds. A FlowAgent receives the intent, evaluates current state, checks policies, and builds a multi step plan to validate capacity, adjust routing, update firewall rules, and open a change record. All of this is reasoning only. When it is time to act, the plan is handed to FlowMCP, which converts each step into validated, contract bound operations that enforce schemas, RBAC, policies, logging, and predictable error handling. If a step falls outside approved boundaries, FlowMCP blocks it. The agent provides the intelligence and decision making, while FlowMCP, via the Itential Platform ensures every action is safe, controlled, and auditable.
VMblog: The announcement says that FlowAI creates a “closed AI-to-Action loop.” What security risks are you specifically preventing by keeping this loop closed, and how does this differ from open integrations with AI systems?
Sprygada: A closed AI to Action loop prevents the biggest risks that occur when AI systems can send unverified instructions directly into operational environments. By forcing every agent to pass through a governed execution layer, FlowAI blocks issues like hallucinated commands, malformed API calls, privilege overreach, and unapproved changes that could impact production networks or cloud systems. Open integrations let AI systems interact with infrastructure without strict validation or policy controls, which exposes organizations to unpredictable behavior and security gaps. FlowAI keeps the loop closed so agents can reason freely, but all actions remain fully validated, permission checked, and logged before anything reaches live infrastructure.
VMblog: Looking at the competitive landscape, how does FlowAI’s approach to agentic orchestration differ from what AI-native workflow platforms are doing? What can FlowAI do that they can’t?
Sprygada: AI native workflow platforms focus on letting models generate tasks and run actions directly, which works for lightweight business processes but not for real infrastructure. They blend reasoning and execution, which means the model has direct influence over what gets changed. That creates risk and makes it impossible to guarantee safety, auditability, or policy compliance at scale.
FlowAI is different because it separates the intelligence layer from the execution layer. Agents can reason, plan, and orchestrate, but every action must pass through a deterministic, governed engine (the Itential Platform) that enforces schemas, permissions, validations, and full audit trails. This lets FlowAI handle high stakes network and cloud operations that AI native platforms cannot safely support.
VMblog: Given that FlowAI is launching as a Private Customer Preview, what are the key learnings or validations Itential is hoping to gain before a broader release, and what do you see as the biggest adoption challenges for enterprises?
Sprygada: In the private preview, we want to validate FlowAI in real hybrid environments and confirm that the separation of reasoning and execution works the way enterprises expect at scale. The focus is on how well agents reason over real data, how reliably FlowMCP enforces safety controls, and how cleanly the system integrates with existing automation, ticketing, and change processes.
The biggest adoption challenges are organizational rather than technical. Enterprises need to build trust in AI driven planning, align security teams around a new governance model, and help operations teams shift from manual workflows to agent supported ones. The preview is designed to surface these patterns early so broader rollout is smooth and predictable.
##
