Introduction
Welcome to an insightful conversation with Dominic Jainy, a seasoned IT professional whose expertise in artificial intelligence, machine learning, and blockchain has shaped innovative solutions across industries. Today, Maison Edwards sits down with Dominic to dive into the complex world of AI agent-to-agent communication protocols. We’ll explore the chaos of competing standards, the historical parallels to past IT challenges, the impact on businesses, and potential paths forward for creating interoperability in this promising field. Let’s uncover why too many standards might mean no standards at all and what the industry can do to break this cycle.
Can you share your perspective on the biggest hurdle facing AI agent-to-agent communication protocols today?
The core issue is the sheer number of competing standards trying to solve what should be a straightforward problem. Every major player and even smaller startups are pushing their own protocols, each with unique flavors and promises. This fragmentation creates a mess where AI agents can’t easily talk to each other unless they’re in the same ecosystem. It’s not just a technical annoyance; it stalls the potential of agentic AI to deliver real value in enterprise settings. We’re seeing a repeat of old IT battles, just with higher stakes because AI is such a transformative technology.
Why do you think having so many standards is such a critical problem for the industry right now?
Too many standards fragment the market and prevent a unified approach. When every vendor or community group pushes their own protocol, there’s no common ground for interoperability. This means developers and businesses waste time and resources trying to bridge gaps or pick a “winner” instead of focusing on innovation. It’s a drag on progress, especially for AI, where seamless communication between agents could unlock powerful workflows and automation. Without consensus, we risk turning a promising field into a patchwork of incompatible systems.
How does this situation with AI protocols remind you of past IT challenges, like those in web services or messaging systems?
It’s eerily similar to what we saw in the ‘90s and early 2000s with things like CORBA, DCOM, and the endless WS-* specifications. Back then, every major tech company wanted to own the standard for distributed computing or web services, and we ended up with bloated, overcomplicated frameworks that few could fully implement. Eventually, simpler solutions like REST and JSON won out because they were practical. Today’s AI protocol wars feel like déjà vu—everyone’s trying to be the next big thing, but we’re just creating noise instead of clarity.
What do you think is driving this explosion of competing standards for AI agent communication at this moment?
It’s a mix of genuine excitement about AI’s potential and strategic business moves. On one hand, the field is still young, so there’s a rush to define how things should work, especially as agentic AI becomes central to enterprise tech. On the other hand, vendors see an opportunity to carve out market share by establishing their protocol as the de facto standard. It’s less about solving a universal problem and more about positioning themselves as gatekeepers. Plus, the hype around AI means there’s funding and attention for anyone who can pitch a new “standard” with a fancy white paper.
How much of this push for new standards do you see as tied to business agendas rather than solving real technical challenges?
I’d say a significant chunk of it is business-driven. Many of these protocols aren’t addressing unique technical needs; they’re often repackaged ways to lock customers into a specific ecosystem. If your AI agents only work seamlessly with one vendor’s tools, you’re more likely to stick with their full suite of products. That’s a classic play we’ve seen before. While there are some genuine efforts to innovate, a lot of the noise is about mindshare and securing future revenue streams rather than pure technical advancement.
Can you walk us through how having multiple standards creates silos for businesses that rely on AI agents?
When standards multiply, you get silos because agents built on one protocol can’t easily communicate with those on another. Imagine a business using one vendor’s AI for customer service and another’s for data analysis—those agents might not be able to share insights or coordinate tasks without custom integration work. It’s like having phones that can’t call each other because they use different networks. These silos isolate systems, limit flexibility, and force companies into tough choices about which ecosystem to commit to, often at the expense of broader innovation.
What kinds of extra costs or workload do companies face when trying to make different AI agents communicate across these silos?
The workload is immense. Companies often have to build or buy translation layers—basically, middleware that converts messages from one protocol to another. That’s not just expensive in terms of development time and software licensing; it also introduces latency and points of failure. Then there’s the ongoing maintenance as protocols evolve or new ones emerge. Beyond that, staff need training to handle these integrations, and there’s the opportunity cost of not focusing on core business goals. It can easily run into hundreds of thousands of dollars for larger organizations, all for something that should be plug-and-play.
How does this fragmentation lead to risks like vendor lock-in, and could you share an example of what that might look like?
Vendor lock-in happens when a business invests heavily in one protocol or ecosystem, making it costly to switch to another. If you’ve built your AI workflows around a specific vendor’s communication standard, moving to a competitor means redoing integrations, retraining staff, and potentially losing data or functionality. For example, imagine a retailer using a particular AI platform for inventory management agents. If those agents only communicate via the vendor’s proprietary protocol, the retailer might be stuck with that vendor’s overpriced updates or subpar tools because switching would mean starting from scratch with a different system.
You’ve referenced historical examples like CORBA and DCOM. What lessons from those earlier standards battles should the IT industry apply to AI agent protocols today?
The biggest lesson is that simplicity and adoption beat complexity and ambition every time. CORBA and DCOM tried to solve every possible problem with intricate designs, but they collapsed under their own weight. Meanwhile, lighter approaches like REST gained traction because they were easy to implement and widely accessible. For AI protocols, we should prioritize a minimal, practical framework that anyone can use, then build on it as real needs emerge. We also need to remember that grassroots adoption, not top-down mandates, often decides the winner—focus on what developers and businesses actually use.
Why do you think the industry keeps repeating the mistake of creating too many competing standards despite these past lessons?
It’s a mix of human nature and market dynamics. There’s always this urge to innovate or control the narrative, especially in hot fields like AI. Companies think they can differentiate themselves by owning a standard, and egos get involved—everyone wants to be the one who defines the future. Plus, the financial incentives are huge; if you can establish dominance early, you lock in customers for years. On top of that, the tech world moves fast, and there’s little time for reflection on history when everyone’s racing to be first. We forget the pain of past failures until we’re knee-deep in the same mess.
What do you mean when you say that multiple standards essentially result in no standards at all?
When you have a dozen or more standards for the same purpose, none of them achieve the critical mass needed to become the universal language everyone relies on. Instead of a shared foundation, you get a fractured landscape where nothing is truly interoperable. It’s like having 20 different electrical plug designs in one country—nothing works together seamlessly, and you’re left with chaos. Without a dominant or widely accepted standard, the industry can’t build the network effect that drives adoption and value. It’s effectively the same as having no standard because there’s no agreement on how to move forward.
How does this lack of a unified standard slow down progress in AI agent technology?
It slows progress by diverting energy from innovation to integration. Developers and researchers spend more time figuring out how to make systems talk to each other than on improving the agents’ capabilities or solving real business problems. It also discourages smaller players or startups from entering the space because the barrier to entry—navigating all these protocols—is so high. Meanwhile, larger enterprises hold off on full adoption because they’re waiting for the dust to settle. The result is a stunted ecosystem where AI agents can’t reach their full potential as collaborative tools.
What impact does this have on businesses or end users who are expecting value from these technologies?
For businesses and end users, it’s incredibly frustrating. They’ve been promised that AI agents will streamline operations, automate complex tasks, and deliver insights, but instead, they’re stuck dealing with compatibility headaches and delayed rollouts. A company might invest in AI only to find that their agents can’t integrate with existing systems or partners’ tools, so the return on investment is nowhere near what was expected. End users, whether employees or customers, experience clunky, disconnected services instead of the seamless interactions they were sold on. It erodes trust in the technology and delays the tangible benefits AI could bring.
You’ve proposed the idea of a ‘minimum viable protocol’ for AI agent communication. Can you explain what that might look like in practice?
A minimum viable protocol would be something dead simple that covers the basics of agent communication without overcomplicating things. Think of something like HTTP paired with JSON, using a handful of standard message types—request, response, notify, error. This would handle the vast majority of interactions between agents, like passing tasks or sharing data, without getting bogged down in edge cases. It would be lightweight, easy to implement, and open for anyone to use, with room to add extensions later as specific needs come up. The goal is to get everyone speaking the same basic language first, then refine it over time.
Why do you believe a simple approach, like HTTP with JSON, could meet most needs for agent communication?
HTTP and JSON are battle-tested and widely understood. They’re already the backbone of most web communication, so there’s a huge pool of tools, talent, and infrastructure to support them. Most agent interactions don’t need fancy protocols—they just need a reliable way to send and receive structured data, which JSON handles well, and a robust transport mechanism, which HTTP provides. By sticking to something familiar and proven, we avoid reinventing the wheel and reduce the learning curve for developers. It’s not about solving every problem out of the gate; it’s about enabling broad adoption and interoperability from day one.
How would starting with a basic protocol help prevent the overcomplicated designs we’re seeing now?
Starting simple forces the industry to focus on core functionality rather than speculative features. Right now, many protocols are bloated with bells and whistles for niche scenarios that might never materialize, which makes them hard to implement and adopt. A basic protocol sets a clear baseline—everyone agrees on the essentials, and we build trust and momentum from there. It also lowers the risk of vendor-driven complexity, where companies pack in proprietary extras to differentiate themselves. By iterating on a minimal foundation, we can add complexity only when it’s justified by real-world use, not hype or agendas.
Looking ahead, what is your forecast for the future of AI agent communication protocols if the industry doesn’t address this issue soon?
If we don’t tackle this fragmentation, I see a future where AI agent technology remains a niche tool rather than a transformative force. We’ll have a handful of dominant vendors with walled gardens, each controlling their own incompatible ecosystems, while smaller innovators struggle to break through. Businesses will face higher costs and slower adoption, and the promise of seamless, collaborative AI agents will stay just out of reach for most. Without a push for interoperability—whether through a minimal protocol or a concerted industry effort—we’re looking at years of wasted potential and another chapter in IT’s history of self-inflicted gridlock.