GLOSSARY
MCP
The Model Context Protocol, or MCP, standardises connections between AI applications and available tools, data sources or templates. It defines how these components communicate. The information an application may read and the actions it may perform still require explicit rules.
Preparing for a customer meeting across systems
Relevant information may sit in a CRM, approved documents and public company websites. Suitable MCP connections can make these resources available through a more consistent interface. This can make integration work reusable, while data quality, permissions and the meaning of individual fields remain separate responsibilities.
The economic question is whether a usable workflow requires less maintenance. Another connection offers limited value if employees still reconcile conflicting records manually. Reading information, preparing a proposed change and writing to a system should therefore be treated as separate capabilities.
Test permissions and failure behaviour
A meaningful acceptance test checks whether a salesperson can access records outside their authorised scope, how missing permissions are handled, and whether a write operation matches its specific approval. A policy engine can support decisions about permitted actions; actual interfaces must enforce them. External content also remains a possible source of prompt injection. A common protocol does not remove that risk.