On my very first day of Seattle Tech Week I discovered people from completely different industries discussing Model Context Protocol (MCP). First, Michael A. Agustin from Curie talked about MCPs enabling retailers to provide product details through AI agents. Then Siyi (Annie) Xue from Brex showed how designers can use MCPs to rapidly prototype in high fidelity without coding, connecting different tools and services seamlessly.
As someone who’s been recently wrestling with the Notion MCP to reliably respond to my own LLM, I was genuinely surprised to see this technical protocol being discussed across retail, finance, travel apps, and product design.
For those unfamiliar: MCPs essentially give large language models specialized abilities by connecting them to external tools and data sources. Instead of trying to build one AI that knows everything, MCPs let you plug in specific capabilities—whether that’s accessing your company’s product catalog, design tools, or customer data. And it’s not just to read data – these allow the invocation of real-world actions and workflows.
What struck me most was Annie’s point about the blurring lines between engineering, design, and product management. When designers can prototype functional experiences by connecting APIs through natural language, and when product managers can directly access real-time data without engineering handoffs, the traditional boundaries start to dissolve. This echoes themes I recently discussed in my prototyping lecture for Pia Zaragoza‘s system design class.
It’s fascinating to see a protocol emerge as a common thread across such diverse industries and roles. The future of work might be less about what role you have and more about how quickly you can connect the right tools to solve the right problems.