Why MCP Matters for Product Managers: My Take on Building Smarter AI Products

How I Got Hooked on MCP

So, picture this: I’m in a sprint planning meeting, pitching an AI assistant for our app that can answer user questions like “What’s my project status?” without them digging through a million tabs. Everyone nods, but then engineering gives me that look—the one that says, “Cool idea, but it’s gonna be a nightmare to hook this AI up to our systems.” We’re talking weeks, maybe months, of custom code to tie it to our CRM, internal APIs, you name it. I leave the room deflated, wondering if I’m asking for the moon.

Fast forward to a random Sunday night, and I’m scrolling tech blogs (yeah, I’m that nerd). I stumble across something called MCP—Model Context Protocol, cooked up by Anthropic in 2024. It’s basically a way to let AI models talk to tools like Slack, GitHub, or your database without jumping through endless hoops. My PM brain lights up. This could be the cheat code to ship features faster and make users happy. Not gonna lie, I stayed up way too late reading about it, and now I’m kinda obsessed.

Why This Matters for PMs

As a PM, what gets me excited about MCP is how it lets us build products that feel effortless. It’s like giving your AI a universal remote to control whatever tools your users already love. Want an AI that pulls data from Salesforce and updates a Jira ticket? MCP makes that doable without a year-long engineering saga.

For me, it’s about solving real problems. Say your users are drowning in notifications—sound familiar? With MCP, you could build an AI that scans their inbox, Slack, and calendar, then prioritizes what matters. No more “I missed that email” excuses. Or imagine a fitness app where the AI checks your workout history and suggests a plan right in the app, pulling data from wearables without a clunky sync process. That’s the kind of seamless stuff that makes users stick around.

A Couple Examples That Got Me Thinking

I’ve been chewing on how MCP could play out in the wild. One’s a real story from a friend, and the other’s a half-baked idea I’ve got scribbled in my notebook.

• Real Talk: A PM I know at a small SaaS startup used MCP to make their onboarding way less painful. Their AI now grabs user data from HubSpot and walks people through setup, like a virtual tour guide. She said it cut their onboarding time in half, which is huge for keeping customers. The best part? She didn’t have to micromanage a ton of integrations—her team just pointed MCP at their tools and let it rip.
• My Crazy Idea: I keep daydreaming about a PM tool with an AI buddy that’s actually useful. Like, you ask, “What’s slowing us down this sprint?” and it checks GitHub for open PRs, Asana for overdue tasks, and Slack for who’s swamped. MCP would be the glue holding that together, so I’m not begging engineers for one-off scripts. Maybe I’ll build it one day… or at least pitch it to someone with more coding chops.

Stuff like this gets me pumped because it’s not just tech for tech’s sake—it’s about making products that save time and feel intuitive.

My PM Game Plan for MCP

I’m no expert, but I’ve been thinking about how to weave MCP into my work without pretending I know Python or whatever. Here’s the approach I’m playing with:

• Figure Out the Win: I start by asking what user headache an AI could fix if it had free rein over our tools. Maybe it’s summarizing customer complaints from Zendesk and Twitter in one go. If it sounds like a game-changer for users, it’s worth exploring.
• Weigh the Cost: MCP takes some upfront work to set up, no question. I’m learning to ask, “Is this a one-time thing, or will it make future features easier?” If it’s reusable, like a pipeline we can tweak for other projects, I’m sold. For scrappy teams, I’d test it on something small first, like one integration.
• Team Up Smart: I lean on my engineers to sort out the nuts and bolts, but I make sure I’m clear on what I need—like, “Can the AI grab X and update Y?” Drawing a quick flowchart of the user experience usually keeps us on the same page.
• Track What Matters: Success isn’t about lines of code; it’s about users. If we launch that AI assistant, I’m watching stuff like how many clicks it saves or whether people use it daily. Good numbers mean we’re onto something.

The Not-So-Fun Parts

MCP’s awesome, but it’s not all rainbows. When I first brought it up, my team wasn’t exactly throwing a parade. Engineers grumbled about learning something new—fair, they’re slammed already. And my director was like, “Can’t we just stick with what we know?” Here’s how I dealt:

• Winning Over Devs: I learned real quick not to oversell MCP as magic. Instead, I showed them it’s not that different from stuff they already use, like APIs, and pointed to ready-made tools online. One dev got curious when I mentioned there’s a community building MCP servers, and now he’s the one hyping it up.
• Convincing the Boss: Higher-ups don’t care about tech jargon—they want results. I pitched MCP as a shortcut to launch features users will rave about, like that onboarding trick my friend pulled off. I also suggested starting with a low-stakes test to avoid freaking anyone out.
• Getting Over the Hump: I’ll be honest, MCP’s docs made my eyes glaze over at first. I’m not diving into code, so I just skimmed enough to get the gist and leaned on a teammate to walk me through a demo. Once I saw it in action, I was like, “Okay, I get why this is cool.”

Where I Think This Is Going

I’ve got a hunch MCP’s part of something bigger. AI’s moving past cute chatbots to agents that can actually do stuff—like handle your expense reports or plan a team sprint. MCP’s the piece that makes those agents feel like they belong in your product, not some bolted-on gimmick. I’m not saying it’s gonna take over the world, but with folks like Anthropic behind it and more tools popping up, I’m betting it’ll be a name we hear a lot.

For us PMs, it’s a chance to push boundaries without burning out our teams. I’m already thinking about how to play with it more—maybe a side project, maybe just bugging my engineers with “what if” questions. Either way, it’s got me excited to build products that don’t just work, but surprise people in a good way.