The Year of AI Agents🔥
How We Got Here and What’s Next
2025 is undeniably the year of AI Agents. And it’s not just me saying that—industry leaders are making it clear:
🗣️ “2025 will be the year of AI agents.” Sam Altman, CEO OpenAI
💡 "Agents are not only going to change how everyone interacts with computers. They’re also going to upend the software industry, bringing about the biggest revolution in computing since we went from typing commands to tapping on icons.” Bill Gates, Co-founder of Microsoft
🌍 "The age of AI Agentics is here, a multi-trillion-dollar opportunity that will redefine how people work." Jensen Huang, CEO Nvidia
But let’s take a step back and understand how we got here.
🤖 What Are AI Agents?
AI Agents are digital entities capable of perceiving their environment, deciding, learning, collaborating, and acting autonomously to achieve specific goals.
Why now? AI has been used for process automation for years, leveraging machine learning and algorithms to streamline manual tasks. However, the introduction of Generative AI has dramatically accelerated progress by enabling AI systems to perform complex tasks with minimal user intervention. Unlike traditional AI, which excels in structured workflows and repetitive tasks, Generative AI uses large language models (LLMs) to generate new content and adapt to dynamic scenarios.
This evolution allows AI agents to assist in complex tasks more autonomously, marking a shift beyond traditional copilot roles towards more independent and creative problem-solving capabilities.
⚡ The Race for AI Agent Dominance
The explosion of AI Agents has led to a highly competitive landscape. Everyone is trying to capture the market:
🔄 Traditional RPA (Robotic Process Automation) players like UiPath are evolving from basic automation to embedding GenAI capabilities.
🌐 Big tech companies (Google, Microsoft, Amazon, Nvidia) and LLM developers (Open AI, Anthropic) are launching AI Agent platforms and integrating them into their ecosystems.
In Europe, Mistral, the AI company specializing in large language models (LLMs), which has raised over $640M at a $6B valuation, is also providing tools to create AI agents through its Le Platform Agent Builder and Agent API. These tools enable the development of sophisticated agents with customizable capabilities, catering to a wide range of applications. You can check it out here.
🚀 New AI-native agent startups are raising billions to build industry-specific or general-purpose AI Agents. In 2024 alone, AI Agent startups raised $3.8B across 162 deals, proving investor confidence in the technology's transformative potential.
There are already hundreds of startups building new products in the AI Agent space, but the potential market is so big, that there’s a lot of room for new players.
Here one of the latest maps on some AI Agents Startups:
🎯 Specialization: Vertical vs. Horizontal AI Agents
The market is moving toward greater specialization, with AI Agents categorized into Horizontal and Vertical Applications.
🔄 Horizontal AI Agents
Horizontal Agents are General-purpose AI solutions applicable across industries. They Provide foundational capabilities in different business areas (Sales, Marketing, Accounting, HR…), that businesses can adapt with customization options.
Today, the two main areas that are having the most adoption for Horizontal Agents are Software Development and Customer Service.
💻 Software Development is being rapidly disrupted by AI, the majority of developers are already using AI-copilots, such as GitHub Copilot, and there is now a huge increase in adoption of Coding AI Agents, some key players leading this transformation include::
Anysphere, the developer of AI-powered coding assistant Cursor that is reportedly in talks to raise capital at $10B valuation
Lovable, a Swedish startup that allows anyone to create apps effortlessly through prompting, that just raised a 15M$ pre-seed round led by Creandum, an that has become one of Europe’s fastest-growing startups, reaching $17M+ ARR in just a few months.
🗣️ Customer Service is another area of major AI Agents adoption, with text and voice AI chatbots providing 24/7 multilingual support and increasingly effective interactions. Key companies in this space include::
Gognigy, a German startup that raised $100M series C round last year and that is already working with industry giants such as Lufthansa, Bosch or Toyota.
Konvo AI, a German/Spanish startup that has built a Conversational AI platform for e-commerce brands that proactively assist customers with human-like conversations that increase conversion.
📊 Vertical AI Agents
Vertical agents are designed for specific industries or domains and they are tailored to address unique challenges within a particular sector. These agents are trained to have deep expertise in a narrow field.
These include AI Agents tailored for various industries:
🏥 Healthcare – Providing AI-powered diagnostics and personalized treatment recommendations. (Example: Brainomix, a spin-out from Oxford University pioneering AI-powered imaging tools, to improve diagnosis and treatment decisions in stroke and lung fibrosis.)
🧪 Biotech – Accelerating drug discovery and molecular research with advanced AI models. (Example: Biorce, startup from Barcelona that has developed the first Clinical AI Assistant for drug development; with its flagship AI model, Jarvis, which aims to streamline data accessibility in clinical trials.)
⚖️ Legal – Streamlining contract analysis, legal research, and case prediction with AI-driven insights. (Example: Luminance, developed by mathematicians from the University of Cambridge, is the most advanced legal LLM today that specializes in automating contract generation and negotiation for legal departments.)
👷 The Rise of AI-Powered "Workers"
AI agents are moving beyond being mere tools and they are starting to take on human-like roles in the workforce.
How AI Agents Function in Work Environments Today:
Analogous Roles: AI agents mimic traditional job roles (e.g., AI SDR in Sales) for easier adoption and are performing specific tasks.
Land and Expand: Organizations are starting to delegate to AI workers a few specialized non-value adding tasks, its use will expand as they improve.
Human Management: Today AI Agents still need human supervision and intervention and there are still concerns on predictability, accuracy, compliance, and regulation.
Usage-Based Pricing: Costs depend on tasks completed or outcomes achieved, not per-seat licenses. This opens some questions on how to predict these costs and how to optimize them with smaller models. It's also important to note that the budgets for implementing AI agents often come from labor replacement savings rather than traditional software expenses.
We are still at the beggining of this Agentic Revolution and there are big discussions on how AI Agents will augment human workers or replace them. I’d love to hear your thoughts on this topic around the future of work! 👥
🔮 What’s Next? Collaborative AI Agents
The next frontier is Collaborative AI Agents that can communicate with other AI Agents, coordinate, and make decisions autonomously. Here 3 examples of companies pushing into this direction:
One of the most exciting developments in recent weeks is Model Context Protocol (MCP), an open standard protocol released by Anthropic. It provides a universal, open standard for connecting AI systems with data sources, replacing fragmented integrations with a single protocol. The result is a simpler, more reliable way to give AI systems access to the data they need. This means AI Agents can now instantly integrate with tools and live data sources, making them more collaborative, powerful and autonomous. If you're interested in a deeper dive, you can check out this video.
There are also appearing new startups in the Orchestration Layer acting as Multi- Agent platforms, such as Crew AI, a Brazilian startup backed by Insight Partners. CrewAI enables enterprises to connect multiple task-specific AI Agents (called ‘Crews’) that communicate with one another, delegate tasks, and automate complex business processes and workflows.
In China, the launch of Manus some days ago, is already shaking up the world. Manus would be the world’s first fully autonomous AI agent; it thinks, plans, and executes independently, thanks to its multi-agent architecture. By deploying specialized sub-agents for different tasks, it can handle complex workflows more efficiently, an ability still uncommon in most autonomous AI systems. Magnus phenomenon has already been described as the "second DeepSeek moment". You can read more about how Manus changes everything, here.
We’re at the starting point of Collaborative Multi-Agent Platforms, and there’s no doubt we’ll see even more groundbreaking examples soon.
I’m both excited and, I must admit, a bit fearful of the radical changes ahead, but I’m optimistic that they will drive innovation and progress. Looking forward to continuing to explore this fascinating evolution!
📩 Want to Learn More?
In upcoming Substacks, I’d like to interview some founders building on AI Agents, so if you’re working in this space or know about really cool AI Agents startups, reach out!
And don’t forget to subscribe to stay ahead in the AI revolution. 🚀