Why MCP and A2A are Key to Solving a Key AI Challenge: Agent Chaining

The enterprise rush towards AI is finally pivoting from broad experimentation towards tangible, reliable results. As organizations look to integrate AI deeper into their workflows, the focus is turning towards systems that offer not just novelty, but genuine efficiency, trustworthiness, and a clear return on investment. The next wave isn't just about smarter individual models; it's about how they work together.

• Chain of events: "Agent chaining is going to be the biggest trend. We'll see a mix of general-purpose chat models for nuanced customer interaction, combined with specialized reasoning models for reliable, accurate decision-making", says Anurag Karuparti, Senior AI Cloud Solution Architect at Microsoft and author of Generative AI for Cloud Solutions. The hybrid model-type approach allows teams to build highly performant, cost-effective, and trustworthy AI solutions that can automate complex enterprise-scale workflows.
• MCP and A2A: But raw model power isn't everything. Connected systems require a re-thinking of infrastructure to match the power of agentic orchestration. Emerging communication protocols like A2A and MCP are top-of-mind for Karuparti as he explores what's possible in AI agent collaboration. "[A2P and MCP] enhance an agent's capabilities by streamlining its access to organizational data resources and tools. They really enhance the performance and collaboration aspect of agents in a way that helps us build better AI applications."
• Finding the right sources to learn from: But adoption of any new tech requires org-wide buy-in and a learning curve that is requiring faster uptake by the day. "At this point, you have to rely on non-traditional sources," Karuparti advises, noting AI's breakneck pace and where traditional education struggles to keep up. "It’s not easy for any company to maintain documentation that reflects real-time updates; it becomes outdated within a week. I educate my customers to rely on credible sources like influencers on X or LinkedIn."

• Leaning into learning: Karuparti’s LinkedIn Learning course demystifies advanced tools and model selection for a broader audience to help augment day-to-day workflows. "The course covers how reasoning models differ from traditional chat models, what kind of applications they truly shine in, how to build cost-effective applications using them, and how they can boost productivity," he says, adding that the latest reasoning models bring more autonomy and can automate large parts of knowledge work.

Karuparti argues that AI proficiency is now table stakes for new entrants to the workforce. "If you're fresh out of college, having the know-how to use all these AI tools is the highest priority skill. And the faster you become more productive compared to your peers, the more you can do with less."