What do AI Product Managers need to know about MCP Servers?

​
The Bridge Between Vision and Execution

Are you building an AI product but spending too much time worrying about integrations? As a product manager or entrepreneur, you've likely experienced this frustration: your brilliant AI concept requires access to various tools and data sources, but each integration becomes its own mini-project.

WAIT!!! I am an Entrepreneur or a Product Manager? Why do I need to know about MCP servers. Isn't my technical team going to figure it all out?

Absolutely but as a business person you need to know what inputs your product can take, access to what resources or tools will make your product valuable and what are the outputs of your product.

Very similar to a homeowner with a large backyard, you need to envision what your backyard will eventually look like, what flowers, fruit and vegetables you want to grow and what seeds you might require. Maybe you will hire a landscaper to help.

Moreover as an Entrepreneur and/or a Product Manager you need to validate your assumptions before going too much into building the product and finding that customers will not buy it. In today's day and age of AI, you don't really need to hire a technical team to validate your assumptions. Learning about MCP servers will greatly open your visibility and access to tools and resources you want to expose in your AI product.

Think of MCP Servers Like Your Mac's USB-C Hub

Imagine you have a MacBook Pro with only USB-C ports, but you need to connect to HDMI displays, USB-A devices, SD cards, and ethernet cables. Rather than building custom cables for each connection (the old API way), you use a single, well-designed adapter hub that handles all these connections for you.

MCP servers function just like that adapter—they're the universal connector between your AI application and the diverse ecosystem of tools and data sources your customers already use. You don't need to understand the complex technical specifications of each connection; you just need to know the hub exists and what it can connect to. And just like Apple's adapter, once connected, data flows seamlessly in both directions with minimal configuration.

Out of the box LLM vs. LLMs integrated with your resources and tools

Traditionally you use the large language model with the data that it has access to. That might include the internet or not but it will not have access to the tools you use.

If now you want the large language model access to your tools and resources in and outside of your network. For example: databases, calendar appointment tools or messaging tools, you had to use a custom api provided by the vendor for each tool and it would be a lot of work creating and maintaining that integration.

MCP servers allow you to easily and seamlessly connect external tools, databases and resources to access data and make it available to your Large Language Model (LLM). If you are building an AI product you need to depend on a LLM and if in your enterprise you have large data sets or external tools that you need to expose to that LLM so it can help you make decisions or take action. MCP servers negate the need to build a custom API to every and / or tool. MCP servers are supported and updated by the vendors of these tools

You just need your LLM to speak to the MCP server for that tool and never worry about any updates to the tool or its API.

How to configure an external tool?

If you are using Claude as an example connecting to any tool and providing it access it just a few clicks as shown in the visual below

​
​

​
​

​
​

How did I experiment with this knowledge on my product?

I have a Jira Telemetry plugin in the Atlassian marketplace - https://marketplace.atlassian.com/apps/1217549/at-a-glance-telemetry-dashboard-for-jira?tab=overview We support both cloud and data platforms.

This is a great tool for software development teams that are looking to enhance feedback loops from tools like Jira, Jenkins, Bitbucket, Github, Travis and so many others. They can hook all of these tools in the plugin and get real time data and metrics as they progress from picking a user story in a sprint and delivering it to customers.

Besides built in charts and data, I wanted to provide my customers with a way to analyze their data using their favorite Generative AI tool.

I experimented with adding a MCP server to my product. I added a chat interface to demonstrate the AI functionality. Trial users get a limited set of queries and paid users get more queries. If my users like to connect directly to the MCP server they get the URL And credentials to get access to their datasets.

Currently our product uses Claude.ai as a tool which can also be accessed with MCP. This is optional (more to demonstrate the use of AI). The valuable resources is the data from my product "At A Glance" and utilize a charting tool to chart the data.

​
​

How can you use MCP Server in your product?

Most vendors have started including MCP server in their products. Check the AI model you use and search for the tools it supports.

MCP server technology can be a powerful tool for product managers to enhance their product's value proposition. Here's how you can leverage it strategically

Strategic Applications for Product Managers

​

1. Automated Customer Insights: Deploy MCP agents to analyze customer feedback across multiple channels simultaneously. Automatically categorize, prioritize, and extract actionable insights from user data to create continuous feedback loops that identify emerging user needs in real-time
2. Enhanced Decision Making Use collaborative agents to run complex scenario analyses for feature prioritization. Implement agent-based A/B testing frameworks that adapt in real-timeCreate simulation environments to predict feature impact before development. This can be very effective for validating your product idea without hiring a technical team
3. Workflow Optimization Automate routine product management tasks while maintaining quality. Deploy specialized agents for documentation, competitive analysis, and market research. Create seamless handoffs between product, design, and engineering teams.

​

What I've Learned Along the Way

My experiments with AI capabilities continue to evolve as I learn more about what truly delivers value to customers. The most important lesson: start small, focus on one valuable connection, and expand based on user feedback.

Your Turn

What's one tool or data source that would transform your product if you could connect it to an AI model without complex integration work? I'd love to hear your thoughts and challenges.

​

Anil Jaising, ​CST®​ On a mission to help Entrepreneurs and Product Leaders THRIVE, Unpack Product Innovation with AI Trainer, Product Consultant and International Speaker Follow me for real life case studies and learning videos.