MCP is driving massive change across the tech industry and if you earn a living in or off of technology, this affects you whether you are technical or not.
In 2024, Anthropic, the company behind Claude, established a new open standard called MCP: Model Context Protocol. What MCP does is connect front-end AI clients like Claude or ChatGPT directly with back-end applications, databases, and systems of record.
Here is why this matters.
We are witnessing the convergence of two powerful forces. On one side: the ability to type free-flowing sentences into a GPT interface and leverage the collective knowledge of large language models. On the other: the accuracy and reliability of back-end systems and data sources that produce factual, deterministic responses.
Traditionally, GPTs produce probabilistic responses, highly accurate but still probabilistic. MCP enables those front-end AI clients to tap directly into back-end applications and databases that return actual facts. All of the benefits of generative AI combined with all of the benefits of established systems. That is a fundamentally different capability.
A Massive Architectural Shift
This rapid evolution in software architecture is arguably bigger than the historical shift from client/server to web. We are effectively shifting back to a client/server model, but with the reach and intelligence of modern AI on the front end. Anyone can now use plugins or connectors to bridge any front-end to any back-end using MCP.
A Warning for SaaS Businesses
If you are in the business of building software or SaaS and you do not offer unique value that cannot be replicated through MCP interoperability, your business model is at risk. The companies that will survive are those that possess high-value proprietary data or unique utility that cannot easily be replaced by connecting a GPT front-end directly to a back-end data source.
Career Advice
If you are evaluating new roles in tech, ask yourself: why would customers do business with this company if a GPT connected directly to a back-end system could replicate what they do? That is the durability test going forward.
Someone said to me recently: “But we built great agents on the front-end for staying power.” That is 2023 thinking – pre-MCP. Agents built by dozens of engineers cannot compete with the collaborative output of millions of users contributing to Claude, ChatGPT, or Gemini. The LLMs will cross-pollinate and catch up.
MCP is changing everything. It may take a few more years to fully play out – but the direction is clear. The software industry needs to be swimming with this tide, not against it.
Want to see MCP in action applied to GTM and revenue execution? Join me for a live 60-minute executive briefing, reserve your seat here.