The adoption of enterprise AI has exposed a fundamental limitation. Large language models (or simply known as LLMs) are excellent at reasoning in conversational interactions, but they have no understanding of how your organisation works.
Imagine your smartest engineer has just joined the organisation. They are smart, capable and eager, but they have no idea how anything works yet. They can follow the playbooks, probably, but they still lack the systemic understanding that only comes with context. They don’t know your operating model (the fundamental how of your organisation), your workflows, the systems or the rules that govern how work moves across teams. They will over time, when they have context. The same is true for even the most powerful AI tools and models.
Model Context Protocol (MCP) is emerging as a popular solution. MCP gives AI assistants a secure and structured way to understand the tools, functions and datasets they can use inside a company’s technology and process landscape. Instead of relying on prompts or heavy fine tuning, which can be very expensive, MCP provides a live, governed interface that allows an assistant to call defined functions, call APIs and execute operational actions with control, traceability and auditability.
This represents a shift from theoretical intelligence to operational intelligence. By exposing system capabilities through a consistent schema, MCP allows an AI assistant to reason about the tools available and decide which one to use based on the task at hand. It bridges the gap between an LLM’s reasoning ability and the reality of how work gets done inside the organisation.
Let’s take a practical example
Imagine an employee messaging is using Teams or Slack and is having a chat with a Halo or ServiceNow agent:
“I cannot access my Sharepoint folder. Can you help me?”
Without MCP
The assistant can only talk about the issue. It will suggest possible causes and offer basic troubleshooting, links to knowledge articles. Indeed, it understands the problem but it has no visibility of the organisation’s access rules, processes, runbooks or service flows. It cannot act because it cannot see the operating model behind the request.
With MCP
The assistant can both understand and resolve the issue. Because platforms like Halo and ServiceNow are exposing their capabilities through MCP, the AI can now:
- Check permissions through the access management API.
- Retrieve the user’s profile and validate their role.
- Query the service management tool to check for related incidents.
- Raise a new ticket prefilled with diagnostics if needed.
- Trigger the correct automation script/runbook to restore access.
- Validate the fix by rechecking the system.
- Respond to the user with:
“Your access has been restored. Try again and let me know if you need anything else.”
All performed within defined governance, aligned to your organisational policies with full auditability.
Why This Matters
This is the difference between advice and action. MCP gives the assistant real understanding of the operational environment, not just the text of the request. It lets the AI behave like someone who understands how the organisation works rather than an outsider guessing from the sidelines.
The Market Is Already Moving In This Direction
ServiceNow was one of the first major platforms to adopt MCP, using it within NowAssist to give AI governed access to workflows, tasks and knowledge artifacts. More recently, Halo has introduced MCP support inside its platform, enabling AI agents to interact with Halo tools including ticket operations, runbooks, reporting and automation sequences.
These capabilities are not bolted-on AI features. They reflect a seismic shift where AI becomes part of the Organisations operational fabric rather than a layer sitting on top of it.
Why MCP Changes the Enterprise AI Landscape
MCP defines how an agent interacts with the system environment. This includes tool descriptions, inputs and outputs, permitted actions, security boundaries, error-handling behaviour and multi step logic. Service providers expose capability once. Then any MCP-compliant assistant, whether accessed through Teams, a portal or a web chat, can then reason about those capabilities and use them safely.
This aligns AI with the organisation’s operating model rather than leaving behaviour to improvised prompts or bespoke integrations. This creates a scalable pattern for AI-driven automation, reduces development and integration overhead and accelerates the Organisations transition from POC/experimentation to real operational impact.
What Comes Next?
MCP is early in its adoption curve in the Enterprise but the direction of travel is clear. As more platforms follow ServiceNow and Halo, AI will shift from being advisory to operational, acting inside systems with control, reliability and a deep understanding of how work gets done.
If you wish to explore how MCP can operationalise and automate how work flows through your organisation, whether on ServiceNow or Halo, drop us a message.
Written by Darren Gerry, Engagment Director
Get in touch with our team to learn more about our services, explore partnership opportunities or discuss how we can help with your challenges.