Synopsis: Dan Fernandez, vice president of product for developer services at Salesforce, explains how vibecoding, in combination with the rise of the model context protocol (MCP), will transform application development.
Vibe coding, Fernandez explained, centers on describing what you want to build and letting AI handle much of the implementation. It can shorten time to value and open development to a wider range of contributors, from product managers to analysts. While this broadens participation, it also introduces risks, particularly when handling sensitive data. In the enterprise, security, governance and policies must be built into the process from the start.
MCP acts like a “USB for services,” enabling AI tools to not only generate code but also take actions — running queries, calling APIs, or producing visualizations. This makes it easier to integrate existing enterprise services and intellectual property, provided role-based access control and compliance requirements are maintained.
Fernandez notes that vibe coding works well for personal or departmental applications, rapid prototypes, and low-priority backlog items that may not require professional developer resources. The key is to define “federal” rules for all applications, along with “state” or departmental rules, so that citizen developers work within secure and compliant boundaries.
Pilots are a common way to introduce the approach: start small, apply guardrails, and refine governance before broader rollout. Done well, this can ease IT backlogs, improve feedback loops, and encourage collaboration between citizen and professional developers.
Looking ahead, Fernandez sees more agent-to-agent interactions and structured inputs expanding what MCP can handle. However, he stressed the need for humans in the loop and explicit controls over how information is shared. With the right framework, these tools could make AI-assisted development a standard part of enterprise workflows.
