Plot Your Path: The 2024 AI Agent Ecosystem Map

The hottest growing space in the AI industry is the AI agent space. Agents are autonomous entities that will complete tasks while humans are sleeping. You may have one agent working for you, or someday, a thousand swarm agents.

The implications are significant. If agents fetch information and conduct tasks on behalf of humans, it means that humans no longer need to visit ad-filled websites, apps with varied user experiences, or fuss with clunky enterprise software. The information comes to humans in a single interface, in the way they want to consume and interact with it. Billion-dollar internet business lines are set to be disrupted—and billion-dollar new business lines are set to emerge.

I’ve been watching this sector for over a year, in June 2023, I published an overview of the AI tech stack, with a nod that agents would be the hot category. In spring 2024, I hosted the largest AI Agent event in the world, Llama Lounge 8. 

To chart the billion (or trillion) dollar new business opportunities, I set out on a research project to map the market. Given my background as an Industry Analyst, I conducted research. Yup. I interviewed the experts in the field, aggregated the discussions, and built a market map. I’m no stranger to this; I did this for Web2, the sharing economy, the tech wellness space, and more.

I’m going to be hosting a private event for AI Agent software entrepreneurs in Nov 2024. You can contact me by sending an email (see contact page) with your company details, and I’ll review it.

Review my five predictions on how agents change the digital world.

---

Overview: The 2024 AI Agent Ecosystem

Agent Ecosystem Layer: Marketplaces

• Foundational Models box (OpenAI GPT as a precursor)
• Enterprise box (Agent.ai by Hubspot, Salesforce AgentForce)
• Big Tech box (Amazon, Google, Microsoft, Meta, Apple)
• Startups box: (MultiOn agent leaderboard)

Agent Application Layer:

• AI Agent Apps box: Largest sector, thousands of companies to emerge. Multimodal: input/sensors, knowledge data, output. No-code agents; in the future: dynamically generated.
• Connect to LLMs box: Most agents are tapping LLMs for NLP but they will also connect to other models for multi-modal capabilities.
• Agent Platforms box: Enable developers to build agents (MultiOn, Adept, Lyzr, CrewAI)

Agent Management Layer:

• Agent Permissions/Security Box:  Agent authentication (KYA credit to Skyfire), Tiered credentials, Agent capabilities.
• Management Box: Orchestration: observability (AgentOps), compliance, swarm management. Arbitration: agent priority in the network.  Payments: agent to tech (Skyfire), agent to human (Payman), human to agent. Improvement: metacognition, reflection (credit to Parth), evaluation, self-healing

Data Layer: 

• Exclusive/Private Data: RAG/Access to enterprise, government, personal data, agent data
• Open Data: public data (Dendrite), data providers, scraping services
• Unified APIs: fast info/transactions, no imitating click paths.

---

Market Predictions and Insights

AI Agent Startups:

• These are the fast movers, and will be the dominant leaders, billion if not trillion dollar companies will emerge out of this category.
• Startups will begin in a single category box, then, as they mature, expand to other boxes.
• Currently, there are hundreds of AI agent startups; this number will expand to thousands by 2025.

Agents from Big Tech: 

• Big tech companies focused on a horizontal market strategy, like Amazon and other cloud services companies, will help startups expand.
• Pristine garden tech companies like Apple will seek to offer a safe and private single Meta AI agent, likely keeping other agents on their marketplace (with a 30% transaction fee) as a paywall.
• Hubspot’s Agent.ai will integrate within its main suite. Salesforce is set to launch AgentForce in mid-September, birthing thousands of enterprise AI agents that initially resemble Copilots or AI Assistants.

Industry Standards: 

• Questions will arise about whether AI agents will have interoperability between these various systems, leading to the emergence of new translation AI agent startups to serve the entire ecosystem.
• A discussion on AI agent standards will ensue, but tech giants will intentionally defend their kingdoms, posing a challenge for true market interoperability.

Impacts to Business and Society:

• Media, advertising, and e-commerce empires will be severely disrupted by AI agents as they bring salient information to humans and conduct tasks on behalf of humans.
• New business models will emerge, involving agents in advertising, e-commerce, affiliates, and more.with significant impacts on business and society starting now.

Most Humans Won’t Comprehend the Growth:

• AI agents will create more AI agents, birthing more agents than the billions of humans within a few years.
• Humans currently spend 7 hours a day online. If the promise of agents (and AI in general) comes to pass, agents will complete high-friction tasks for us, allowing us to be more human.
• AI agents will be the dominant entity on the internet (and within businesses), This will happen at a very rapid pace as in 2-3 years agents will create agents, at an exponential rate, human minds will not be able to comprehend.

This map will be updated on a periodic basis. I look forward to speaking at a conference near you to give the market overview, and a forecast of what’s to come.

---

Interviews with: Chris Yeh, Div Garg, Omar Shaya, Sarah Allali, Scott Johnson, Jeff Abbott, Parth, Craig DeWitt, Amir Sarhangi, Mikhil Raja, Alex Reibman, Christopher Kauffman, David Schatsky, Chris Saad, Charles Maddock, Ori Neidich, Siva Surendira, Matt Schlicht, Ben Parr, João Moura, and more.