AI: High Time we move from OS to Agents

In a recent interview, Jensen Huang, CEO of NVIDIA, made an assertion: the last decade has been about building a new operating system - an AI OS - and that the LLM is it. And that the next many more decades will be about building on top of this OS. An OS that of course will continue to be rapidly enhanced.

But LLMs are only the latest evolution in a much broader and deeper AI-driven OS "ecosystem". While LLMs have garnered much attention, we must recognise that the AI field has been steadily contributing components to this new operating system for decades. Beyond Generative AI, there are critical innovations in Text-to-Speech (TTS), Speech-to-Text (STT), Optical Character Recognition (OCR), and Document Processing. These technologies have been quietly but profoundly enhancing enterprise workflows for at least the past 15 years.

These components, together with more recent breakthroughs, form a complete ecosystem of a nascent, but very competent, AI OS.

The only problem is.. everyone.. thinks the "OS" will solve every use case imaginable.

While that might be true for some use cases, enterprises have struggled to adapt this OS layer to practical everyday needs and workflows.

And all of this is.. because of the missing "agentic" layer.

Read on to know why that is the next wave here to stay for many more than the next 10 and how it is Agentic and Autonomous. #FourthIndustrialRevolution

The Building Blocks of the AI OS

AI-powered systems now encompass a vast array of technologies. Each has advanced independently but is now coalescing into an ecosystem of tools and functionalities that work together to solve complex problems. Consider just a few of these advances:

• Machine Learning and Data Science: Machine learning algorithms and the field of data science have been critical in analysing vast amounts of structured and unstructured data to help detect patterns, to forecast trends, and inform decision-making with the power of predictive insights.
• OCR and Document Processing: Companies have been improving document recognition and processing capabilities for years. These tools enable enterprises to automate mundane document workflows, from invoicing to regulatory filings, reducing human error and saving time.
• Text-to-Speech and Speech-to-Text: Advances in natural language processing (NLP) and speech synthesis, such as those from Google and Amazon, have transformed customer service and accessibility solutions. From virtual assistants to real-time transcription, these tools help bridge the gap between humans and machines.
• Image Processing and Video Generation: AI-powered image recognition and generation tools are now capable of creating high-quality images and videos from textual descriptions, enabling use cases from marketing to entertainment.
• Generative AI (LLMs): While LLMs are among the most visible examples of AI innovation, they represent only one piece of a much larger ecosystem. Yes, they are the "OS" in some sense—allowing for natural language interfaces and transforming data into actionable insights—but they are but only one "OS component".

Beyond the Hype: What Enterprises Truly Need

While the perceived "hype" of Generative AI is deserved - given the ease with which these models can now generate text, images, and even code - the reality is that enterprises have struggled to adapt these AI OS layers to their daily operational needs. The expectation that an AI "OS" will seamlessly address every use case imaginable has fallen short.

This disconnect arises from a misunderstanding of AI's role in business processes. Most enterprises need AI not to create poetry or generate art, but to work autonomously, methodically solving specific problems within workflows. While AI can certainly generate new opportunities, it’s the unglamorous tasks - the grunt work - that are crying out for automation.

The Missing Layer: Agents for Every Workflow

This is where the next stage of AI innovation comes into play: AI agents. Enterprises do not just need an intelligent "OS" - they need autonomous agents that are focused, specialised, and purpose-built to get real "work" done within defined workflows. These agents would serve as the workhorses of AI, handling the "unsexy" but essential tasks like data extraction, decision-making, and multi-system integration.

The concept of an AI agent is not new. In fact, it mirrors the role of human employees in a workflow - experts who manage a specialised set of tasks, but with one key difference: AI agents can operate autonomously. They can infer, decide, and act based on real-time data, making them uniquely positioned to handle the increasingly complex demands of modern enterprises.

For example:

- In Financial Services, an AI agent could autonomously process loan applications by pulling data from disparate systems, verifying information, and flagging inconsistencies for human review.

- In Healthcare, agents could assist clinicians by analysing medical records, running diagnostic algorithms, and even suggesting treatment plans based on patient data and historical outcomes.

- In Manufacturing, agents could oversee entire supply chain processes, making real-time decisions based on current stock levels, transport availability, and shifting demand forecasts.

- In Advertising, AI agents could autonomously manage entire advertising campaigns, from initial ad creation to budget allocation to audience targeting to continuous analysis and optimisation for engagement and conversions.

Agent Specialisation: The Key to Unlocking Enterprise AI

Each agent must be task-specific. Just as an enterprise might employ specialised teams for finance, operations, and customer support, AI agents must be tailored to the precise needs of each workflow. These agents need to be "made for" the particular use case in question, embodying the expertise that would otherwise require significant human effort.

Moreover, these agents must be capable of interfacing with multiple systems, learning from data streams, and adapting to changing conditions. The agentic layer will therefore need to be flexible, constantly evolving to incorporate new AI capabilities, and continuously integrating with enterprise systems and workflows.

We are entering a new phase of AI adoption, where the focus shifts from the marvel of what AI can generate to what AI can do - concretely, reliably, and at scale. This shift will require a new focus on developing autonomous, agentic systems capable of acting on their own to fulfil specialised tasks in real-world business environments.

The Future: From AI OS to Fully Agentic Ecosystems

The vision for the future is clear: businesses will no longer rely on a general-purpose AI "OS" to address every need. Instead, they will deploy fully autonomous agents across their use cases, creating an AI-driven ecosystem where each agent is specialised, intelligent, and able to operate independently within its task directive.

The possibilities are endless. Think of a future where:

• Sales agents autonomously prioritise leads, engage with potential customers via personalised AI-generated content, and schedule follow-ups based on predictive analytics.
• HR agents manage employee records, conduct preliminary interviews, and suggest career development plans based on historical performance data.
• Customer support agents autonomously resolve issues, escalate cases to human operators only when truly necessary, and maintain a detailed log of interactions for compliance purposes.

This agent-driven future will allow businesses to scale like never before, applying AI to tasks that were previously too complex or labor-intensive to automate. The time for experimentation is over - enterprises need to build and deploy AI agents that can autonomously operate within their workflows maximising potential of the AI OS Ecosystem.

---

AS: This article has been expanded upon with the help of AI. ????