MCP Multiplies the Usefulness of Incumbent Products

A new interface layer is emerging for software.

For most of the history of software, the primary interface was the user interface. Then it was the API. Now, increasingly, software is also becoming something AI systems can use directly.

That matters most for incumbent products.

In vertical markets, incumbents already possess what is hardest to build: trust, distribution, embedded workflows, and domain knowledge. In insurance, healthcare, energy, or public administration, customers do not replace core systems lightly.

So the important question is not whether AI will replace these systems. More often, it will make the best ones more useful.

MCP is one way this happens.

What changes

Model Context Protocol gives AI applications a standard way to interact with software. In practice, it means a product can expose a set of capabilities that agents and workflows can discover and use directly.

That changes the role of the product. Its usefulness is no longer confined to its own interface.

A system that was previously used only by a person at a keyboard can now participate in a broader chain of work. The product becomes more valuable because it becomes easier to use in context.

Buying decisions and timing

As AI workflows become more common, customers will increasingly prefer systems that fit easily into them. A product that is simple to connect gains an advantage over one that is closed, even if both are otherwise competent.

There is also a timing effect. The first companies to make this interface available, and to document it clearly, are more likely to be discovered, recommended, and adopted. Once a product becomes the obvious system to connect, that position reinforces itself.

The reverse is also true. Products that remain closed may continue to serve existing customers well, but they begin to look narrower than they are. Over time, that becomes a disadvantage.

A more sensible AI strategy

For many companies, MCP is also the more sensible AI strategy when it comes to cost.

Building proprietary in-app AI requires infrastructure, model decisions, ongoing cost management, and constant maintenance. An MCP layer avoids much of that. Customers can use the models and tools they already prefer. The vendor makes the product accessible without having to own the entire stack.

That is usually the better trade.

The work itself is not complicated in principle. Expose useful tools. Tie them to existing permissions. Price them in proportion to usage and value. Write public documentation that is clear enough for both developers and models to understand.

Incumbents' first step in the AI era does not have to be to become something else. They can already create a lot of value by extending what they already are.

MCP multiplies existing products' usefulness.