When people first hear about Model Context Protocol (MCP), they think of it as a simple technical gateway for an AI model to talk to an API. That’s true, but it’s the smallest part of the story.
At Propel, we see MCPs as organisational infrastructure. They’re becoming the rails that AI systems run on. And if you only treat them as backend plumbing, you’ll miss some of the biggest opportunities: revenue, productivity, and customer experience.
Most teams design MCPs with developers in mind. But once you add natural language interfaces, something changes: non-technical users step in.
We’ve watched analysts, paralegals, and operations staff use MCP-powered workflows to pull data, update records, or draft documents—without touching an API. Suddenly, it’s not just about developer experience. It’s about how your whole organisation interacts with AI.
When MCPs are treated as small side projects, we see the same traps:
Lost revenue: SaaS companies that don’t connect MCPs to their monetisation strategy miss chances to differentiate, upsell, or advertise capabilities.
Becoming relics: MCPs built for one integration drift into irrelevance. Without ownership, metrics, and evolution, they’re technically correct but strategically disconnected.
Forgetting the user’s user: In one client project, we refactored complex workflows into natural language. The big win wasn’t just better staff UX—it transformed the experience of their customers’ customers.
AI maturity isn’t about whether you “have an MCP.” It’s about how embedded MCPs are across your business. Ask yourself:
Are they tied to critical workflows?
Do they evolve alongside your product strategy?
Can non-technical staff use them easily?
Do you measure ROI and usage like you would any other product surface?
Well-designed MCPs act like a second brain. They connect systems, cut across silos, and surface context where it’s needed. AI agents handle the repetitive pulling and reconciling so people can focus on higher-value work.
We’ve seen teams save hours every week by collapsing multi-step processes into one natural language request. This isn’t theory - it’s happening now.
The real opportunity is to treat MCPs as part of your business model:
Monetisation: How you capture and signal value.
Product roadmaps: How you build capabilities that users discover and adopt.
Org design: How non-technical staff interact with AI day to day.
If you want AI to stick, MCPs need to be more than connectors. They should shape pricing, new roles, and customer experiences. That’s how you move from technical plumbing to building the rails of an AI-native future.