Infrastructure
What MCP changes for business software—and what it does not.
Model Context Protocol can standardize how AI systems connect to tools and data. Here is the business case without the protocol hype.
October 12, 2026 · 6 min read
The short version
MCP can reduce bespoke integration work, but permissions, data quality, tool design, and operational ownership still determine whether the system is useful.
02
Resources, tools, and prompts are different contracts.
MCP distinguishes between information an application can read, actions it can invoke, and reusable interaction patterns. That separation encourages clearer system boundaries. Reading a customer record is not the same kind of capability as issuing a refund, and the permission model should reflect that difference.
This is valuable architecture, but the protocol does not magically clean your data or make an unreliable API dependable. It standardizes the connection; it does not replace the work of defining access, validating inputs, logging actions, and handling failure.
03
Where MCP creates immediate business value.
MCP is most useful when a company has several systems that people repeatedly consult while making decisions. Sales teams might need CRM history, proposals, call notes, and product availability. Operations teams might need ticket status, inventory, policy documents, and finance data. A common access layer can make these sources available without copying everything into one new platform.
- —Internal assistants that need governed access to several sources.
- —Agent workflows that must discover and use tools consistently.
- —Teams maintaining the same integration for more than one AI interface.
- —Organizations that want to separate data access from model choice.
04
Treat every MCP server as a product surface.
A useful MCP implementation starts with a narrow business job. Tool names need to be unambiguous, returned data should be concise, and high-impact actions need explicit approval. The operational owner should know which systems are exposed, who can access them, and what is recorded for audit.
The strongest first project is rarely ‘connect everything.’ It is one valuable workflow with clear read and write boundaries. Once that works reliably, the same connection pattern can expand across the organization.
Continue reading
Related analysis
AI strategy
AI agents or workflow automation? Choose the simpler system first.
A practical decision framework for choosing between deterministic automation, AI-assisted workflows, and autonomous agents.
Applied AI
Context engineering is the difference between a clever demo and useful AI.
Why business AI needs the right information, tools, memory, and boundaries—not one enormous prompt.