The Dawn of Agentic AI - Part 1
Exploring the Next Frontier of Artificial Intelligence: Unleashing the Power of Independent AI Agents
Welcome to the first in an exciting new series of articles exploring the cutting-edge world of Agentic AI. Over the coming weeks, we'll dive into the current state, future potential, and emerging trends in this transformative field of artificial intelligence. As investors and innovators in the tech space, it's crucial to understand how Agentic AI is poised to reshape industries and create new opportunities.
Understanding Agentic AIs
Agentic AI, also known as autonomous AI, represents the next frontier in artificial intelligence. Unlike traditional AI systems that simply respond to specific inputs, Agentic AI aims to create AI agents capable of independent decision-making and task execution. Think of it as AI that can take initiative, adapt to new situations, and work towards achieving goals with minimal human intervention.
Key Components of Agentic AI
To better understand Agentic AI, let's break down its core elements:
Foundation Models:
At the heart of Agentic AI are Large Language Models (LLMs). Think of these as the "brain" of the AI agent. They're like incredibly advanced text prediction engines that have been trained on vast amounts of data from the internet. This training allows them to understand and generate human-like text, reason about complex topics, and even solve problems. It's similar to how a human expert gains knowledge and analytical skills through years of study and experience.Tool Use:
This is the AI's ability to interact with the digital world around it. Imagine giving your AI assistant the power to use your computer or smartphone. It can open applications, search the web, send emails, or even make changes to documents – all on its own. This capability allows the AI to not just give advice, but to actually perform tasks and access information as needed.Memory:
Just like humans, AI agents need a way to remember information and learn from past experiences. This isn't just about storing data; it's about understanding context over time. For example, an AI with good memory can recall previous conversations with you, remember your preferences, and apply lessons learned from past tasks to new situations. This makes the AI more personalized and efficient over time.Multimodal Capabilities:
This means the ability to work with different types of information. While early AI mostly dealt with text, Agentic AI aims to understand and create various forms of data. This includes interpreting images (like recognizing objects in a photo), processing speech (understanding spoken commands), and even generating visual or audio content. It's like giving the AI the full range of human senses and expressive abilities.
The Agentic AI Landscape
The field of Agentic AI is rapidly evolving, with various approaches and applications emerging:
Horizontal Applications of Agentic AI: These include general-purpose tasks like customer support and sales automation that work across industries.
Vertical Applications of Agentic AI: Industry-specific solutions for sectors like healthcare, finance, or manufacturing.
Agent Infrastructure: Platforms and tools for building and deploying AI agents.
Two Paths to Agentic AI
As the race to develop powerful AI agents heats up, companies are broadly taking one of two approaches:
1. Model-First Startups
These companies believe that the AI model itself is the most critical component in creating effective AI agents. They're investing heavily in developing more advanced language models, betting that significant improvements in AI capabilities will come from this foundational technology.
Key players in this space include:
OpenAI (Over $13 billion raised): Known for GPT models, they're reportedly working on a personal work assistant that can control a user's computer. The highly anticipated GPT-5 is expected to launch later this year.
Anthropic (Over $7.3 billion raised): Following a similar strategy to OpenAI, but at a smaller scale. While they haven't made public announcements about AI agents, industry insiders suggest they're actively researching in this area.
Adept ($413 million raised): Taking a unique approach by training models on user actions, learning from how people interact with their web browsers.
Imbue ($220 million raised) and Magic AI ($145 million raised): Both focusing on AI agents for software engineering, developing their own specialized models.
The key question these startups are trying to answer is: What type of model is best suited for creating truly effective AI agents? Is it a super-powerful general model like GPT-5, a user-action model like Adept's approach, or a reasoning- and code-focused model like those from Imbue and Magic? Time will only tell, as more companies are saturating the landscape.
2. Workflow Applications
In contrast to the model-first approach, these companies are leveraging existing AI models and focusing on how to apply them effectively to specific tasks or industries. They believe that the "glue" connecting AI to real-world applications and user interfaces will be the key differentiator.
We can view these companies on a spectrum:
Vertical Task Automation: These companies create AI agents for specific industries. For example, Harvey is developing AI agents for the legal sector.
Specific Task Automation: Companies like Cognition Labs focus on one complex task (in their case, writing code) that applies across multiple industries.
Horizontal AI Agent Platforms: At the far end of the spectrum are companies like Lindy, which offer platforms with multiple AI agents capable of handling a variety of tasks like scheduling, note-taking, and document summarization.
The Promise of Agentic AI
While we're still in the very early days, the potential of Agentic AI is immense:
Enhanced Productivity: AI agents could handle complex, multi-step tasks, freeing up human workers for higher-level thinking and creativity.
Personalized Assistance: Imagine having an AI assistant that truly understands your needs and can proactively help across various aspects of your work and life.
Innovation Catalyst: Agentic AI could accelerate research and development across industries by autonomously exploring solutions and running experiments.
Improved Decision-Making: By processing vast amounts of data and identifying patterns, Agentic AI could provide valuable insights to inform strategic decisions.
Looking Ahead
As we conclude this introduction to Agentic AI, it's clear that we're on the cusp of a seismic shift.
In our upcoming articles, we'll dive deeper into specific applications of Agentic AI, explore the technologies driving its development, and discuss how startups and investors can position themselves to capitalize on this emerging trend.
At Untapped Ventures, we're excited about the potential of Agentic AI to create new markets and transform existing industries. We're actively seeking out and supporting innovative startups working in this space, as we believe they'll play a crucial role in shaping the future of AI and its impact on business and society.
Stay tuned for our next installment, where we'll explore real-world use cases of Agentic AI and how they're already beginning to reshape various industries.
As shared by Andrew Ng, “AI agent workflows will drive massive AI progress this year—perhaps even more than the next generation of foundation models. This is an important trend, and I urge everyone who works in AI to pay attention to it.”