Model Context Protocol: How To Integrate It Into Your Product Strategy

Forbes

2 hours ago

Ashay Satav is a Product leader at eBay, specializing in products in AI, APIs and platform space across Fintech, SaaS and e-commerce.
Model context protocol (MCP) has been the talk of the town lately, and it's emerging as a foundational part of the AI ecosystem. From enterprises launching their MCP servers to startups developing tools and platforms that support MCP inclusion, the tech world is increasingly gearing up to welcome this protocol, which will accelerate their AI journey. Since Anthropic open-sourced MCP in late 2024, it's played a crucial role as a standard for sharing enterprise data with AI agents, which can then connect with multiple databases and services without custom, pricey integrations.
It's not just a growing fad, as major vendors like Postman, Alation and Cloudflare have announced the incorporation of this protocol in their offerings. This ever-increasing support indicates that many tools and platforms will speak the standard in the future. With decade-long experience in building and commercializing API products at companies like Blackhawk Network, Intuit and eBay, and having witnessed how interconnectivity creates value, I firmly believe that MCP will be key for AI to create real customer impact. For instance, consider a small business that sells products on eBay and uses QuickBooks to manage its accounting. They can use a chatbot that utilizes MCP to securely connect with their eBay and QuickBooks accounts and query "What were my top-selling products last month, and how are my profit margins?" thereby building a strategy tailored to their specific needs. As I draw on my experience to share my thoughts in this article, I believe MCP offers product managers (PMs) a unique opportunity to increase their products' adoption by participating in the broader AI ecosystem.
Integrating MCP Into Your Product Strategy
Embracing MCP is as much a strategic decision (for interoperability) as a technical one. Below are some high-level guidelines and considerations for product managers looking to leverage MCP in their roadmap.
One should identify areas where a richer AI context can solve clear problems or add value, such as customer support, personalization or analytics. Prioritize use cases where integrating AI with live data addresses known pain points. For instance, an MCP-powered AI assistant accessing your knowledge base can significantly improve the user experience. The product manager (PM), as always, should prioritize the highest impact areas over incremental advances.
The MCP ecosystem and adoption are growing exponentially. PMs can start utilizing existing MCP integrations with your current or planned software tools. Many CRM, file storage and ticketing services already support MCP. If you're a platform provider, consider creating an MCP server to facilitate easier integration with third-party AI services or agent deployments. This will make your platform more valuable, as other enterprises will also consider using the "plug-and-play" approach.
Leverage MCP as a governance layer to ensure the safe development of AI innovation. The PM should involve security and compliance teams early, configuring MCP's robust security controls (e.g., consent, access roles and audit logs) to their needs. From here, clearly define what data and actions AI can access. The great thing about MCP is that these guardrails apply centrally.
Above all, given that this is a new standard with high potential impact, plan to educate your team or build the right team with an experimentation focus. Encourage teams to prototype with MCP in low-risk scenarios first, such as internal pilots, to build experience and refine integration based on feedback. This will motivate the team to learn fast and launch with confidence.
Challenges To Adoption And Mitigations
Although MCP promises a novel vision of interconnected AI agents, implementing MCP in a typical enterprise system can present some challenges.
Enterprises must carefully choose the content and context they feed to AI models, as they have limited context windows. MCP provides a standardized approach to managing and prioritizing this context. Techniques include summarizing older data, compressing less essential details and retrieving relevant information when needed. Implementing a "context engine" alongside MCP, such as a vector store or rule-based filtering, helps manage context efficiently. Although a technical feature, the PM should consider incorporating this as part of the strategy.
Adopting MCP involves managing integration changes to avoid breaking AI application interactions. PMs should treat them like APIs, with versioning and backward compatibility in mind. This helps ensure stable interfaces, even if underlying services change, such as switching from Zendesk to ServiceNow.
Wide access to AI raises security and privacy issues for enterprise data. MCP lets organizations enforce governance directly at integration points, such as anonymizing sensitive data or excluding confidential information. PMs should involve security teams early, implement strict access controls via tokens and maintain detailed logs of AI queries and data responses. Future MCP solutions could include advanced consent-based sharing and audit capabilities.
Implementing MCP requires new technical and product team skills, including backend integration, prompt engineering and data governance. During onboarding new teammates and addressing the learning curve, PMs should ensure that pilot projects, community resources and training with practical examples are involved. Start with a few read-only integrations in testing environments and then gradually expand capabilities to meet scale.
Conclusion
The model context protocol may be new, but it's quickly becoming a multiplier for enterprise AI product strategies. By investing in MCP, organizations are laying the groundwork for a wave of AI innovation that's context-aware, personalized and deeply integrated into business operations. Product managers should begin treating context as a first-class citizen in AI projects. This involves asking vendors about their context strategy, planning how different systems will feed into your AI's understanding and ensuring you have governance for this new interaction layer.
In the race to AI-enabled transformation, MCP offers a robust and scalable strategy to fuel the adoption of AI agents that are not only intelligent but truly integrated into your business.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?