Are you investing today to build yesterday’s information architecture tomorrow? The Model Context Protocol (MCP) is not just for AI tools but changes the way we think about data and data products in the enterprise.
The Problem: Nobody wants your dashboard 🧭
I speak with enterprise leaders and teams who are busy upgrading their organization’s data management and building dashboards. We live in a data economy, after all. They are integrating data sources, building data data platforms using Databricks, Snowflake, and similar architectures, and the best of them bring clarity to data provenance and are developing robust data dictionaries. These dictionaries are then used by developers who are producing dashboards and other data products for the organization.
Meanwhile, everyone in the enterprise are busy using conversational AI tools like ChatGPT, Copilot, Gemini, and so on.
Bad news: nobody wants a dashboard anymore. They never did. They want answers or inspiration. Dashboards were the best tools we had.
The Action: Adopt the right AI-First approach ✨
We absolutely still need to build high-quality, well-tested, robust data products for our organizations. The popular idea that AI will make all hard analysis go away is just wrong. Please continue to build great models and compelling visualizations.
But the AI as the preferred interface – that idea I can get behind. Just look around on your colleagues’ screens. Does more than half of them have an AI tool open as an app or in search?
Recall the “mobile-first” paradigm in user interface architecture, where we recognized the futility of forcing desktop interfaces onto mobile devices and instead embraced designing specifically for the mobile as the primary interaction point.
Now, we are entering the era of the AI-first paradigm.
Conversation may not be the only user interface, but it is a formidable one that is here to stay. The AI as an orchestration engine (‘reasoning model’) is powerful and is already transforming our productivity and helping us discover new depths of insights.
The enabling technology for AI orchestration is the Model Context Protocol (MCP). This allows your AI engine (“MCP Host”) to use services (“MCP Server”) to gather additional information to use as context for the query. This can be anything: databases, file systems, models, forecasts, external data, ….
Your data product must become an MCP Server. It can also be a dashboard, an Excel plugin, a database, a complete application, or something else. But it must be an MCP server. Its capability as an MCP Server will define your new enterprise integration layer.
This is the baseline expectation. Without it, your information architecture will be obsolete in 18 months because it will not support the orchestration layer nor your users’ preferred interface.
The Impact: New skills, new architecture ⚡
Some things stay the same, some things change. Your data dictionaries will now be available through MCP so they need to serve not just developers but also end users by describing what the data means in the language your colleagues and customers use and understand.
Your developers had data and user interface (UI) skills. You still need data and modelling, and you may still need UI skills. But you absolutely need API skills so your teams can create compelling, useful, and robust MCP servers.
A crucial evolution for your enterprise will be the development of an MCP Server that acts as a comprehensive registry or repository for other MCP Servers. This is likely the second MCP Server you want to build. As the enterprise is moving beyond traditional dashboards and leveraging AI as an orchestration engine, a robust discovery mechanism for data and context is absolutely crucial.
This approach creates a flexible, extensible data fabric where new “models of context” can be added by various teams, yet remain discoverable and usable across the entire enterprise, without creating silos or requiring manual, ad-hoc connections.
There are some interesting challenges to solve around managing roles with a chat interface. For example: “Am I asking this question in a personal capacity or as the firm’s HR analyst who needs unrestricted access to sensitive data?”
This is not fundamentally different from role management in dashboards or other data products but a comprehensive and intuitive user experience model for this scenario is still actively evolving.
The Bottom Line: Start today 📌
You need to start today. AI adoption is the fastest technology adoption we have ever seen. Ignore the hype about AI taking away all the jobs or eliminating all the thinking. Focus on AI as interface and as orchestration. Dig deep into what it means for you and for your organization.
Are you still building yesterday’s information systems?
About me: I help organisations turn complex data into clear decisions and commercial outcomes. My focus is on enabling better decision-making and unlocking new value through data-driven innovation – especially where the stakes are high and the problems are difficult and poorly defined.
Follow me on LinkedIn for more insights.