Module 2

The Build / Buy / Ignore Decision

Should your team actually invest in MCP? A decision framework.

---

The Scenario

You have read Module 1. You understand the case for MCP. Now the hard part: should your team build an MCP integration, buy a managed solution, or deliberately ignore it for now?

Your engineering lead says MCP would take two sprints. Your designer says users are not asking for it. Your data team says competitors are shipping it. Your CEO wants a recommendation.

This module gives you a framework to make that decision with confidence, backed by data and business logic instead of hope or panic.

---

The MCP Relevance Test

Not every product needs MCP. But the five questions below will tell you if yours does. Answer honestly. There are no wrong answers, only answers that lead to different decisions.

Question 1: Do Your Users Also Use AI Assistants?

This is the foundation. If your users are not already using Claude, ChatGPT, Gemini, or other AI assistants, MCP integration has limited immediate value. That said, the growth curve is steep. In 2024, enterprise usage of AI assistants was 30 percent. By 2025, it was 65 percent. By mid-2026, it is heading toward 90 percent in knowledge-worker categories.

The question is not "are they using AI now" but "will they be using AI in the next 12 to 18 months." The answer is almost certainly yes.

Signal: If more than 20 percent of your user base is already mentioning AI assistants in feature requests or support conversations, you have high relevance. If zero percent are, you have low relevance, but low relevance is a temporary state.

Question 2: Does Your Product Have Data or Actions That AI Agents Would Want?

This is a filter. Not all products are useful to AI agents. A game has limited appeal to an agent workflow. A messaging app does too. But a project management tool, a CRM, a spreadsheet platform, an analytics dashboard, a code repository, or an inventory system? Those are all packed with data and actions that agents want.

If your product stores, analyses, or acts on structured information that would be valuable in an agent workflow, you have relevant functionality. If your product is primarily consumed, not used, you do not.

Signal: List your top 20 API endpoints. If more than half of them would be useful to an agent (reading data, creating records, updating status, triggering actions), you have high relevance. If less than a quarter would be useful, you have low relevance.

Question 3: Are You Losing Deals Because Competitors Have AI Integrations?

This is the accelerator. Even if your users are not explicitly asking for MCP, are you seeing customer churn or lost deals because a competitor has an AI integration and you do not? If so, urgency is high.

Ask your sales team. Have they lost deals explicitly because of missing AI integrations? Are prospects asking "does your product work with Claude or ChatGPT?" If the answer is yes, this becomes urgent.

Signal: If your sales team reports even one lost deal per quarter due to missing AI integrations, urgency is high. If they report none, urgency is lower.

Question 4: Do You Currently Maintain 3+ Custom API Integrations?

This is the efficiency play. Every custom integration is maintenance debt. If you are already maintaining integrations with Slack, Zapier, and a handful of other platforms, MCP becomes a cost-saving opportunity, not just a feature opportunity.

Signal: Count your custom integrations. If you have three or more, MCP is a way to consolidate and reduce maintenance burden. If you have one or two, the savings are lower.

Question 5: Is "AI-Native" Part of Your Product Positioning?

This is the strategic question. Some products are building themselves explicitly around AI integration (agents, AI-powered workflows, agentic features). For those products, MCP is table stakes, not optional.

Signal: Does your product roadmap mention AI? Does your marketing talk about AI capabilities? Does your CEO mention AI in investor updates? If yes to all three, MCP is strategic. If no to all three, MCP is optional for now.

---

Interpreting Your Answers

Score yourself on each question. If you have 4-5 high relevance signals, build MCP now. If you have 2-3 high relevance signals, build within the next two quarters. If you have 0-1 high relevance signals, ignore for now but revisit quarterly.

---

The Build vs. Buy vs. Ignore Matrix

Once you know your relevance score, the decision framework is straightforward.

When to Build vs. Buy

Build in-house if: You have a mid-to-senior engineer available for 2-4 weeks. You have clear visibility into which product features are most valuable to agents. You want full control over what gets exposed and how it evolves.

Buy a managed solution if: You do not have engineering bandwidth available. You want someone else to maintain the server infrastructure. You are okay with a more generic exposure of your product's capabilities. Managed MCP providers (like Anthropic's MCP server hosting partners) can have you live in days instead of weeks.

Hybrid approach: Many teams build the initial MCP server in-house (to ensure it reflects your specific product), then move to managed hosting later once the scope is stable.

When to Ignore (For Now)

If your answers to the five questions tell you to ignore MCP for now, that is a valid decision. But make it deliberately, not by default.

Set a quarterly trigger: "We will revisit MCP if [metric X] hits [threshold Y]." This keeps it on your radar without distracting from current priorities. Possible triggers include: 10 percent of sales conversations mention AI, a competitor launches MCP, your CEO explicitly asks about it again, or you are already spending time on custom integrations.

---

First-Mover vs. Fast-Follower Strategy

The assumption in many organisations is that fast-follower is always smarter than first-mover. Let someone else take the risk. Learn from their mistakes. Ship better.

That logic does not apply to protocol shifts. When an entire platform ecosystem standardises on a single protocol, the early movers get disproportionate distribution. Mobile apps in 2008. Browser extensions in 2012. API integrations in 2016.

The companies that shipped first got the halo of being in the App Store early. They got top placement. They captured user mindshare. By the time fast-followers arrived, the early movers had momentum.

With MCP, being fast-follower instead of first-mover costs you 6 to 12 months of ecosystem visibility and acquisition. That is not trivial.

If you are in a category where MCP makes sense, be first if you can. If you cannot, be fast.

---

The Cost of Waiting

One more framework. Waiting on MCP has three concrete costs:

1. Lost ecosystem visibility. Every day that passes, AI assistants are being asked "do you know a tool for X?" If your product is not on that list, a competitor is. That traffic is not coming back.

2. Recruitment and retention pain. Developers and product people want to work on products that are AI-native. If your team is not building for this layer, you lose candidates to companies that are.

3. Strategic positioning erosion. If MCP becomes table stakes in your category (and it will), your product becomes a legacy play faster. Your TAM shrinks relative to AI-native competitors.

These costs are real, even if they are not on your balance sheet.