The MCP Revolution and the Search for Durable Enterprise AI Use Cases

When Anthropic introduced the Model Context Protocol in late 2024 — an open standard for connecting language models to external tools, databases and APIs — few expected how quickly the initiative would reshape the enterprise AI landscape. A little over a year later, MCP has gone from a niche protocol to a de facto standard supported by dozens of platforms. But behind the rapid growth lies an uncomfortable truth: the industry still has no answer to a fundamental question — which enterprise AI use cases are actually durable?

That is the subject of a long interview AI researcher Sebastian Valkötter gave to KDnuggets. His position is both optimistic and soberingly honest. On one hand, MCP solves a real engineering pain. Before the protocol, every integration of an AI model with a corporate system — whether a CRM, a knowledge base or an internal tool — required custom development. It resembled the pre-USB era, when every device needed its own connector. MCP offered a unified "plug": a standardized way to describe available tools, pass context and receive results. For developers, this means an MCP server written once for Salesforce or PostgreSQL works with any compatible model — Claude, GPT, Gemini or open alternatives.

However, Valkötter points to something the industry prefers to discuss in lowered voices: the security of the MCP ecosystem remains a serious problem. By its nature, the protocol widens the attack surface. When a language model gains access to corporate systems through MCP servers, a whole range of threats appears — from prompt injections that make the model perform unintended actions, to compromised MCP servers that may be swapped for malicious versions. The problem is compounded by an ecosystem that grows faster than auditing and verification practices take shape. The community has already recorded cases where third-party MCP servers contained vulnerabilities that allowed access to data the model should never have touched.

But the researcher's most provocative thesis concerns business logic rather than technology. Valkötter puts the question bluntly: which enterprise AI use cases can be considered stable at all? By stability he means not just technical reliability but durable economic value — a situation where an AI solution predictably earns more than it costs to deploy and maintain. Here the picture is less rosy than vendors paint it. Many companies that deployed AI agents to automate workflows found that the models behave unpredictably in edge cases, need constant human oversight and cost more than expected. The gap between impressive demos and everyday productive work remains significant.

This gap is especially visible with so-called AI agents — autonomous systems able to carry out chains of actions without constant human involvement. MCP essentially provides their infrastructure foundation, giving agents standardized access to tools. But the more autonomy an agent gets, the higher the stakes. An error in text generation is an annoyance. An error by an agent that sends emails to customers on a company's behalf, changes database records or initiates financial transactions is a potential disaster. Valkötter stresses that the industry needs not only technical standards like MCP but also trust standards: formalized ways to determine what level of autonomy is acceptable for a given scenario and what guarantees the system must provide.

The paradox of the current moment is that standardization both accelerates progress and exposes problems. MCP lowers the barrier to building AI integrations, which inevitably attracts more experimenters and increases the number of deployments. But each new deployment is another stress test that reveals the limits of the models, the protocol and corporate readiness. In a sense, MCP plays the same role for the AI industry that HTTP played for the early web: it creates a common language but does not guarantee the quality of what is said in it.

For the Russian market, these conclusions are especially relevant. Domestic companies developing their own language models and AI platforms face a choice: follow MCP as a global standard or develop alternative approaches to integration. Given that the protocol is open and not tied to a specific vendor, supporting it looks like the pragmatic decision — but with investment in their own security and auditing mechanisms.

The main lesson from Valkötter's observations is simple but important: standardizing infrastructure is a necessary but insufficient condition for the AI industry to mature. Until questions of predictability, security and economic viability are resolved, the MCP revolution will remain a revolution of plumbing — impressive engineering whose value is determined by what ultimately flows through the pipes.