A decentralised ‘Internet of Agents’ can provide an alternative to the Big Tech view of our AI-driven future, argues Dr Richard Blythman, co-founder of Naptha.AI.
So much of the discussion around agentic artificial intelligence (AI) depicts a dystopian reality where ‘bots run wild’, reducing the control and decision-making power we should have over many aspects of our personal and professional lives.
I believe much of that fear and scepticism stems from one underlying issue – the vision of AI that has been presented to the public has been almost exclusively that of Big Tech. The public’s understanding of AI has been shaped by companies whose business models rely on centralising data and using it to sell you products.
It doesn’t need to be this way.
Personal AI agents, by contrast, can be self-hosted and preserve the user’s right to privacy. In this way, they are underpinned by a level of autonomy and alignment with the immediate needs of the user to deliver a far more desirable and user-centric approach.
Here in Ireland, in particular, a strong and passionate talent base of agentic AI developers has the potential to help steer the global AI conversation in a way that benefits individuals, small businesses, as well as global giants alike.
Paradigm shift
Rather than just answering questions or generating responses, agentic AI systems are a promising paradigm shift in AI development, offering several advantages over traditional single-agent models.
They are goal-driven; in other words, they plan, act and learn. Agents use tools and can collaborate with other agents to solve complex tasks. However, to date, there has been an interoperability issue – each tech company builds their own agent which exists within its own walled garden.
On Google’s website, I can interact with the Gemini model or on WhatsApp I can interact with the Llama model. Recently, open protocols have emerged like model context protocol (MCP) – an open protocol that standardises how applications provide context to LLMs – and Agent2Agent (A2A), which facilitates the use of a range of online tools. In short, MCP and A2A aim to do for artificial intelligence what HTTP did for the web, which is to establish a lightweight, universal standard for communication. When widely adopted, MCP, as well as other open-source protocols, will allow AI agents to interact with one another seamlessly, regardless of the platform or vendor behind them.
I believe a vendor-agnostic, open and decentralised platform has the potential to change people’s perception of AI.
Instead of agents working for Big Tech and knowing everything about you or your company, the agent remains under your control and acts on your terms alone. The agent can still interact with other agents in the network to optimise resource utilisation, enhance adaptability and improve robustness by ensuring that if one agent fails, others can continue functioning or take over tasks.
This is the vision for the ‘Internet of Agents’, a platform that maintains many of the benefits of agentic interactions while mitigating many risks.
Empowering small businesses
Contrary to the prevailing Big Tech approach, agentic AI that is operating on behalf of a small business can provide significant benefits.
Historically, large online markets have tilted the playing field in favour of larger brands with bigger budgets, funnelling customers towards large sellers. However, AI agents enable organisations to do more with less.
Imagine a small, Irish business that is operating nationwide. With a tailored AI agent, they could automate supplier communications and analyse global trends, all without needing to feed sensitive data into third-party systems or compete for algorithmic visibility against massive multinationals.
In another example, an Irish salon or clinic could use an agent to automate a follow-up to customer bookings with recommended products and after-care solutions.
While there is still a competitive element for algorithmic visibility via smart use of online ads, social media and SEO, the possibilities and applications are endless, but they are also entirely on each company’s own terms. As before, customer data protection underpins every step.
Just as social media gave small businesses a voice two decades ago, agentic AI and the Internet of Agents can give them an operational advantage, if it is deployed strategically. The key difference is that this time, we can build it so that the technology doesn’t simultaneously extract value from users.
Decentralisation
At the heart of this approach is a commitment to decentralisation and open-source development.
When AI is designed to run locally, or on infrastructure you choose, it becomes more secure, more trustworthy and more adaptable.
An agent doesn’t need to send every query to the cloud or siphon your data to advertisers. Instead, it operates within the boundaries you define, learning from your behaviour, serving your goals and staying within your guardrails.
This contrasts with the current dominant AI platforms, which centralise intelligence, aggregate user data at massive scale and serve commercial goals that often diverge from those of individual users.
The Internet of Agents represents a redistribution of influence in the AI ecosystem, and is a critical step to ensure that agentic AI benefits the many, not the few.
Ireland’s opportunity
Ireland is well-placed to lead in this space. We have a pool of impressive and ambitious technical talent, a globally connected start-up ecosystem and an openness to digital innovation.
Hackathons and developer meetups are gaining momentum, and the growing base of active contributors is forcing everyone to think outside the box. Attending these events, I have seen Irish-based teams build sophisticated and capable AI agents within a matter of hours. As a nation, we should be encouraging and creating frameworks for these developers to go further – but there are obstacles.
The EU AI Act risks overregulating at the wrong level. Licensing constraints on models like Llama 4 already limit the tools available to European developers, leading some founders to establish their business overseas as a result. If Europe wants to be competitive in this next wave of AI, and Ireland wants to be more than an attractive hub for importing and housing Big Tech, it must enable responsible innovation rather than restricting companies based on scale.
At its core, AI is truly a neutral technology, and I believe in agents that are transparent, personal and safe by design.
It is true that there is some suspicion and scepticism around agentic AI, but by focusing on what decentralised AI can do for us instead of the other way around, we can have the Internet of Agents that truly benefits us all.
By Dr Richard Blythman
Dr Richard Blythman is co-founder and chief scientific officer with Naptha, a modular AI platform that enables developers to build applications, conduct research and scale cooperative AI agents on the agentic web.
Don’t miss out on the knowledge you need to succeed. Sign up for the Daily Brief, Silicon Republic’s digest of need-to-know sci-tech news.
www.siliconrepublic.com (Article Sourced Website)
#agentic #pave #type #internet