MCP Course Home → Module 2 of 7

The MCP conversation is on the table. Your engineering lead says it would take 2 sprints. Your designer says users are not asking for it. Your data team says competitors are shipping it. Your CFO says "what's the ROI?"

You need to make a recommendation. Build it, buy a managed solution, or deliberately ignore it for now. The wrong call costs you either engineering cycles you cannot spare or competitive ground you cannot recover.

Before you debate resource allocation, answer these five questions. They determine whether MCP even belongs in the conversation for your product.

Question 1: Do your users also use AI assistants?

Check your user research. Are your customers using Claude, ChatGPT, Copilot, or similar tools in their daily workflows? If you do not know, ask. Run a quick survey. Check your support tickets for mentions of AI tools. If fewer than 10 percent of your users are actively using AI assistants today, MCP is probably not urgent. Revisit in 6 months. The adoption curve is steep.

If the number is above 25 percent, keep reading.

Question 2: Does your product have data or actions that AI agents would want?

Think about what your product does. Could an AI assistant meaningfully use your product on behalf of a user? If you build a CRM, the answer is obviously yes: an AI could look up contacts, log activities, generate reports. If you build a niche hardware calibration tool with no data layer, the answer is no.

The test: imagine a user saying to their AI assistant, "Do [thing] in [your product]." If that sentence makes sense, your product has MCP potential.

Question 3: Are you losing deals because competitors have AI integrations?

This is the urgency signal. If prospects are choosing competitors specifically because those competitors work with AI tools and you do not, MCP is not a roadmap item. It is a revenue problem. Treat it accordingly.

Question 4: Do you currently maintain 3 or more custom API integrations for different AI platforms?

If you have already built individual integrations with Claude, ChatGPT, and one or two others, MCP consolidates all of them. This is a cost-saving play: replace N integrations with one. Calculate the engineering hours you spend maintaining those integrations per quarter. That number is your MCP business case.

Question 5: Is "AI-native" part of your product positioning?

If you market yourself as an AI-first or AI-powered product, your customers expect you to play well with the broader AI ecosystem. Not having MCP support is like claiming to be "cloud-native" in 2018 but not having an API. It undermines the positioning.

Count your "yes" answers:

For those of you in the 2-3 range, timing matters as much as the decision itself. Use this quadrant to determine your approach.

The two axes:

X-axis: User demand for AI integration. This is observable. Are customers asking for it? Are competitors shipping it? Is your sales team hearing about it? Low demand is not zero demand. It means "not yet."

Y-axis: Strategic value of ecosystem position. This is forward-looking. How much does being in the MCP ecosystem matter for your product's future? Products that distribute through integrations (SaaS tools, developer tools, data platforms) score high. Products with captive user bases and no integration layer score low.

The four quadrants:

Top-left: INVEST EARLY (High strategic value, low user demand). Your users are not asking for AI integrations yet, but your product's long-term value depends on ecosystem position. Build MCP now while there is less competition for attention in the MCP directories. By the time demand materialises, you will already be established. This is the "plant the flag" quadrant.

Top-right: BUILD NOW (High strategic value, high user demand). You are leaving money on the table every sprint you wait. Users want AI integrations. Your product benefits from ecosystem position. This is the highest priority quadrant. Move to Module 3 immediately.

Bottom-left: IGNORE (FOR NOW) (Low strategic value, low user demand). MCP does not move the needle for your product right now. But do not just ignore it blindly. Set a trigger metric: "When X percent of support tickets mention AI tools" or "When our top 3 competitors have MCP integrations." Revisit quarterly.

Bottom-right: FAST-FOLLOW (Low strategic value, high user demand). Your users want AI integrations, but MCP is not strategically transformative for your business. You do not need to pioneer. But you do need to ship within 6 months of your category leader. If you wait longer than that, you lose the "we have this too" conversation with prospects.

Once you know your timing, you need to decide your approach.

Build your own MCP server when: your product has unique data or actions that no off-the-shelf solution can expose, you have engineering capacity, and you want full control over what is exposed and how. This is the right choice for most products with any meaningful feature set.

Use a managed MCP platform when: you want MCP exposure quickly and your product already has a well-documented API. Managed platforms like Smithery can help accelerate deployment. This is the right choice for products that need speed over customisation.

Deliberately ignore MCP when: your answers to the Relevance Test came back 0-1, and your trigger metrics are not firing. This is a valid, defensible choice. But document it. Write down why you are passing and what would change your mind. The worst outcome is ignoring MCP by default rather than by decision.

Create a simple scoring spreadsheet:

This becomes a living document. Revisit it quarterly. The answers change as your market evolves.

← Module 1: The 60-Second MCP Pitch | Course Home | Next Module: Anatomy of an MCP Integration →