Do any of these commands sound familiar?
- Computer, estimate arrival time.
- Computer, identify the vessel.
- Data, give me a list of anything that was affected by the star’s destruction.
Hopefully, that last command was a clue as to where you may have heard them before. Diehard and maybe even not-so-diehard fans of Star Trek: The Next Generation may recognize these commands given by Captain Picard to Enterprise computers or Data. Yet when recorded in the late 1980’s, human conversations with AI-like systems represented Sci-fi fantasy, not reality.
AI-generated illustration inspired by characters from Star Trek: The Next Generation. Not affiliated with or endorsed by Paramount.
Fast forward to today and conversations between administrators and their IT infrastructure have already started to occur. Further, more infrastructure providers will announce some form of natural-language option for administrators to interact with their IT infrastructure.
Let’s Talk
Recent vendor announcements and third-party publications highlight this emerging trend of administrators conversing with their AI-enabled infrastructures. Consider:
- HPE has introduced Compute Copilot within HPE Compute Ops Management platform. While still in beta, administrators may ask it natural language questions about server inventories, health, firmware levels, and operational status. This eliminates the need for administrators to navigate dashboards and menus.
- Cisco announced its AgenticOps strategy and associated tools. These allow administrators to create workflows and operational actions using natural language instructions. Cisco envisions a future where humans and AI agents jointly manage enterprise networks.
- Selector AI developed a Network Language Model (NLM) and Copilot interface specifically for network and infrastructure operations. Their platform also allows operators to ask questions in plain English and receive correlated telemetry, root-cause analysis, and remediation guidance.
- Microsoft’s Copilot strategy already allows organizations to ask questions about documents, emails, meetings, and business data through conversational interfaces. Microsoft executives openly describe conversational AI as becoming the next major interface paradigm.
Two factors currently drive these initiatives to adopt and implement conversational AI into IT infrastructure.
First, the cost of implementing conversational AI has dropped significantly in the past few years. One study found that performance costs dropped from $20 per million tokens to $0.07 per million tokens, a 280-fold cost reduction. Second, organizations implementing conversational AI often experience significant productivity gains, especially among newer workers. It reduces training time as less skilled and experienced workers may achieve productivity gains of up to 35 percent. Conversely, more experienced workers tend to use conversational AI to fundamentally reengineer existing workflows and make them more efficient.
AI-generated illustration inspired by characters from Star Trek: The Next Generation. Not affiliated with or endorsed by Paramount.
Implementing Conversational AI Simple in Principle
Just because the technology now exists for administrators to talk with IT infrastructure does not mean that this happens auto-magically. However, part of the beauty of how any technology provider can implement conversational AI stems from the simplicity in its architecture.
Conversational IT operations will likely depend on organizations putting the following two foundational capabilities in place:
- AI agents that can reason and act across infrastructure
- Standardized interfaces such as Model Context Protocol (MCP) that allow those agents to securely access enterprise tools, telemetry, and workflows.
Agentic AI enables systems to move beyond answering questions to planning and executing multi-step tasks across multiple systems. That capability becomes increasingly important as organizations shift from “chatting about their infrastructure” to “operating infrastructure through conversation.”
Similarly, MCP standardizes how AI models discover and invoke tools and external systems. The Model Context Protocol project describes MCP as analogous to a “USB-C for AI”. MCP reduces the need for custom integration between models and enterprise systems. MCP can already connect to local data sources such as files or databases, tools such as search engines, and workflows. MCP simply provides a standardized way for AI applications to connect to external systems.
This standardized approach should eventually serve to minimize or eliminate the need for humans to operate product management GUIs. Rather, AI agents will manage and operate IT infrastructures using APIs.
Agentic AI must now only learn the MCP protocol for Agentic AI to work with the APIs from many IT infrastructure providers. Infrastructure providers then only need to expose their product’s capabilities via APIs through MCP adapters.
Technology providers as well as banking and financial companies already see the fundamental shift occurring in IT infrastructure management. Technology providers that include Anthropic, Cisco, Cloudflare, Google Cloud, Microsoft, OpenAI, and Red Hat, among others, already have implemented this approach as part of their respective solution. Further, companies in the banking and financial sector have disclosed plans to invest in Agentic AI and MCP. These include Lloyds Banking Group, NatWest, and RobinHood.
MCP’s Competitive Differentiator
There exists at least one more reason organizations should want to adopt MCP. Organizations complain all the time about vendor lock-in. Yet, inevitably, when it comes time to refresh their IT infrastructure, they often buy from the same provider.
This decision often stems from familiarity. They can manage the vendor’s existing and new platforms using their current management interface(s). This negates the need to learn a new GUI.
MCP helps eliminate this historical barrier to adopting new products and providers. Both emerging and existing technology providers have struggled to gain a foothold in new accounts due to concerns about managing their platform. Using Agentic AI and MCP, organizations get increased flexibility to adopt new platforms and providers. Further, the better each provider implements its MCP provider, the easier it becomes for organizations to justify acquiring that provider’s platform.
Consider VergeIO VergeOS which has already done an MCP provider implementation. In using AI to evaluate VergeOS based upon its available documentation across eight dimensions, VergeIO already differentiates itself from other platforms in the following ways.
| MCP Evaluation Dimension | VergeOS Assessment | Supporting Evidence |
| Breadth of Coverage | Strong | Compute, storage, networking, and data protection capabilities appear accessible through a unified API and CLI. |
| Semantic Richness | Moderate to Strong | Agent skills encode expert procedures and workflows rather than exposing only raw API calls. |
| Discoverability | Likely Strong | The MCP server appears to expose both environment information and platform documentation to AI clients. |
| Permission Granularity | Moderate | Actions appear administrator-gated and API keys inherit user permissions and roles. |
| Explainability | Moderate | Agent skills and documentation grounding suggest explainability, though public evidence remains limited. |
| Failure Impact | Moderate | Human approval gates exist, but rollback, simulation, and dry-run capabilities have not been publicly documented. |
| Cross-Domain Context | Very Strong | A single code base spans compute, storage, networking, and protection within one object model. |
| Auditability | Unknown | Public documentation does not currently describe AI audit trails, reasoning logs, or action histories. |
If anything, evaluating platforms across these eight dimensions may give organizations new impetus to adopt alternative platforms. After all, why use platforms that do not lay the foundation for them to use conversational AI when such platforms already exist?
MCP Levels the Infrastructure Playing Field
Platforms that support Agentic AI and expose robust MCP providers deliver more than conversational interfaces. They enable AI agents to discover capabilities, correlate telemetry, automate workflows, and manage infrastructure using a common protocol. As conversational operations mature, organizations will increasingly evaluate platforms based on the completeness, security, and intelligence of their MCP implementations.
Historically, management GUIs created operational lock-in by forcing administrators to learn vendor-specific interfaces. MCP changes this dynamic by enabling AI agents to manage infrastructure through standardized interfaces rather than proprietary consoles. These place providers on more equal footing and simplify management across heterogeneous environments spanning multiple vendors and technology domains.
FAQ – Do MCP provider implementations vary?
Yes. Just because products support MCP does not equate to providers exposing the same functionality through their respective MCP providers. MCP gives providers a wide degree of latitude in terms of how they expose their platform’s features. Organizations should examine MCP implementations across at least the following eight dimensions:
- Breadth of coverage. Some platforms may only expose read-only information such as a list of volumes or total capacity. Strong MCP implementations will give organizations options to perform actions on the platform. These actions may include creating snapshots, initiating failovers, provisioning storage, or running diagnostics.
- Semantic richness. Agentic AI needs more than APIs, it needs context. Good MCP implementations expose rich metadata that describe the constraints, dependencies, relationships, and risks associated with performing an act.
- Discoverability. Good MCP implementations permit Agentic AI to dynamically discover a platform’s capabilities, such as its ability to create snapshots. Poor implementations typically require hardcoding these features into the Agentic AI.
- Permission granularity. This represents an opportunity and a risk in running Agentic AI. Only giving agents administrative permission to perform all tasks on a platform can constitute an unnecessary risk. Limiting agents to only performing actions based upon assigned, specific roles represent a more robust MCP provider implementation.
- Explainability. Agentic AI that says it completed an action may sound good on the surface. However, good MCP implementation will confirm it completed the action, prove it completed it, and provide rollback instructions.
- Failure impact. Infrastructure operations fail all the time for various reasons. Poor MCP implementations may only report that an operation failed with no explanation as to why. Robust MCP implementations may perform dry runs, validate the procedure beforehand, estimate downtime if the operation fails, and identify affected systems.
- Cross domain context. Organizations typically fail to comprehend both the technical complexity of their IT infrastructure and the dependencies that exist. Agentic AI knowing volume and LUN mappings certainly help document the complexities and dependencies that exist within IT infrastructures. However, understanding the applications and business services that rely upon these technical mappings helps organizations design more resilient IT infrastructures.
- Auditablity. Smart organizations often perform root cause analysis to prevent incidents from recurring. Completing this analysis requires a thorough recording of all the activities that occurred so an organization may audit them. Robust MCP implementations will possess this functionality as it gives organizations more confidence to deploy Agentic AI.
KEEP UP TO DATE WITH DCIG
To be notified of new DCIG articles, reports, and webinars, sign up for DCIG’s free weekly Newsletter.
To learn about DCIG’s future research and publications, see the DCIG Editorial Calendar.
Technology providers interested in licensing DCIG TOP 5 reports or having DCIG produce custom reports on their behalf, please contact DCIG for more information.
Editor’s note: Verge.io is a client of DCIG