"Ogres are like onions."
‍

Most of us know the quote.
‍

In Shrek, the point wasn't really about onions. It was about layers. Things that seem simple on the surface are often much more interesting once you understand what's underneath.
‍

That's becoming increasingly true of the AI solution stack and platform in Maximo Application Suite.
‍

Most conversations about AI in Maximo focus on the Maximo Assistant because it's the part users can see. But the Assistant is only the outer layer of a larger architecture that includes AI Service, watsonx.ai, the new MCP Server introduced in MAS 9.2, and ultimately Maximo Manage itself.
‍

Understanding those layers helps explain both how AI works today, where IBM appears to be heading and how you can leverage the platform when developing solutions. The platform is similar to automation script, once you understand what is there the possibility are nearly endless.  

Layer 1: Maximo Assistant

The Assistant is the part everyone sees. Chat windows have become the default way that we interact with AI. That is why IBM highlights this feature in Manage too. It provides a conversational interface that allows users to ask questions, retrieve information, and interact with Maximo using natural language.  
‍

For many organizations, this will be their first AI experience in MAS. It is easy to understand, explain and justify the funding for. While it will likely be the first way users use the AI platform it is only it isn’t the only way. The MAS AI platform has much more to offer.  

Layer 2: Maximo AI Service

Behind the Assistant sits Maximo AI Service.
‍

This is the platform layer that connects Maximo applications to watsonx.ai and enables AI functionality across the suite. The Assistant is only one feature that relies on AI Service.
‍

Other examples include:

• Problem code recommendations

• Field value recommendations

• Similar work order identification

• Reliability Strategy recommendations

• Lease abstraction in Maximo Real Estate and Facilities
  ‍

AI Service should be viewed as platform infrastructure, not just a chatbot dependency.

Layer 3: watsonx.ai

Beneath AI Service sits watsonx.ai.
‍

This is where the large language models live. At a high level watsonx.ai provides the intelligence. The chatbot isn't the intelligence. When users interact with the Maximo Assistant, requests are ultimately processed through models accessed via AI Service. IBM's architecture describes AI Service sending requests to LLMs hosted through watsonx.ai while supervisory AI agents coordinate the overall interaction.

Layer 4: Maximo MCP Server

With MAS 9.2, IBM introduced the Maximo MCP Server.
‍

This may ultimately be the most important layer in the stack. The MCP Server exposes Maximo capabilities as tools that AI agents can discover and invoke.
‍

I think about the REST APIs as being designed for developers. MCP was designed for AI agents.
‍

Sure, AI could interact with the REST API, I have seen it done. The question you must ask yourself is if that is the best approach. The MCP Server provides a governed way for AI systems to interact with Manage without exposing the entire application surface.
‍

If you want to know more bout the MCP server, my colleague has already published a great, detailed blog post.

Why MCP Matters

The Maximo Assistant is one AI consumer. The MCP Server enables many AI consumers.
‍

Today that might be the Maximo Assistant.
‍

Tomorrow it could be:

• Microsoft Copilot

• Enterprise AI assistants

• Custom copilots

• Agentic workflows

• Third-party AI platforms
  

The MCP Server transforms Maximo from an application that users interact with into a platform that AI agents can interact with. That's a much bigger architectural shift than simply adding a chatbot.

Where Should You Start?

Looking for a turnkey experience?

Implement Maximo Assistant and AI Service. This is the fastest way to put AI in front of users.

Already have an enterprise AI strategy?

Get to MAS 9.2 and explore the MCP Server. If your organization already has AI agents or copilots, MCP may be the most interesting capability in the release.

Looking for IBM's AI-powered features?

Implement AI Service. Most of IBM's current AI capabilities rely on it, including problem code recommendations, work order recommendations, lease abstraction, and Maximo Assistant.

Final Thoughts

The Maximo Assistant gets the attention because it's the layer users can see and understand quickly, it demos well. The deeper story is the architecture underneath it. The MCP Server provides a standardized way for AI agents to interact with Maximo while remaining the system where the work actually happens and keeping control of the data.
‍

Which brings us back to Shrek. The most interesting part of an onion isn't the outer layer. It's understanding how all the layers work together.
‍

The same is becoming true for AI in Maximo. The chatbot may be what users are asking for today, but the more transformative story may be happening deeper in the stack.