The age of AI agents is here.

Very soon, there won’t be many tasks left that AI agents can’t do well. Existing and new frameworks will allow us to build them faster, and more efficiently. Key implementation problems will be solved. While it seems like building the agents is the hardest task right now – it is already getting easier.

Soon, there will be a race to build the key platforms where AI Agents transact, specialize and grow. It’s going to coalesce into a trend – the rise of the AI Agent marketplace.

Marketplaces will be a dominant way people interact with AI agents in the next few years.

Here’s what we’re seeing on the edge of the map:

The AI Agent Strategy Matrix: Where Do You Fit?

To understand why we believe AI agents will organize into marketplaces, we have to go back to the old software days. Back then, the big debate was what angle you should take on the market: horizontal or vertical? There were strong opinions on both sides.

At first, everyone was chasing the big horizontal software opportunities. The angle was broad product = big market opportunity. Then, the tactics flipped, and founders began creating vertical solutions for specific industries — the angle was capturing more of the user experience and owning a whole vertical.

We ended up with both strategies. (Everyone was a little bit right). The key was choosing which angle worked best for you. And there was a great formula I used to think about it: evaluate the complexity of your task and the versus size of your potential customer.

After the initial wave of huge horizontal software companies, something interesting happened. Software with more “broad” solutions — like accounting, or calendar management — tended to see more success going after SMBs. Smaller customers turned out to be a winning formula.

For these customers, the selling point was speed, ease, and a capacity that was simply out of reach before. Generic business processes, boiled down into an easy-to-implement package, was a legitimate value add for these SMB customers.

We saw companies like: Quickbooks, Calendly, or Square. The playbook: Simple or generic task -> go small, and go for scale.

On the other hand, we saw a handful of companies build vertical SaaS to really own the experience of an entire industry. These companies offered help with very specific, complex tasks. And interestingly, they ended up being a better fit for larger enterprises.

Why? These enterprises had the budget to actually pay for unique solutions.

We saw companies like Procore, Veeva Systems, or OpenGov. Their playbook: Complex task -> go deep, get a few big contracts.

Those were two “green areas of opportunity.” Of course there were exceptions, but most companies tended to focus there.

You could visualize this dynamic like this:

So what does this have to do with AI Agents?

First, we’re seeing this pattern repeat itself. Which means there are a few ways to think about the way AI agents are currently being deployed and organized.

First there’s the horizontal use case: using agents to create instant, cheap, and scalable BPOs. Think: agents for accounting, marketing, sales, that can be applied in a variety of industry contexts.

The other way is the vertical use case: AI employees that act like specialists. They are “hired” by companies to specialize in tasks like industry data analysis, legal writing…it’s like hiring an expert at a fraction of the price. We’ve even seen teams of AI scientists trained to test and develop drugs.

Like we saw with software, there’s room for both of these AI agent angles. But we’re already seeing this dynamic unfold in an an interesting way: horizontal AI agent marketplaces for SMBs.

Right now, nearly half of American economic output comes from SMBs. These millions of businesses – from your local restaurant to the neighborhood accounting firm – that represent $750B in IT spending. Yet they’ve been chronically underserved by traditional technology waves. Many still struggle with basic operations, unable to access the expertise and tools that larger enterprises take for granted.

This is where AI agents become truly transformative. For the first time, we’re seeing technology that can democratize access to sophisticated business services at low prices, at scale. The horizontal angle is looking really good for AI agents.

Why? Because of the adoption curve of new technologies. SMBs are going to lead the charge of AI Agent adoption because they have everything to gain. These businesses can’t always afford to hire a whole marketing agency or an accountant. But they already know the benefits would be 10x on day 1. Agents take away the parts of running your business that you can’t do — or hate to do. It’s an easy sell.

We’re already seeing the success of this model. Our portfolio company Triple Whale —  a comprehensive dashboard that integrates all the tools needed build an online DTC business —  has already layered AI agents into their system. You can choose to activate an agent to perform a task you may constantly overlook or not have time for. In this case, those agents are built right into the software, and it’s an instant lift for e-commerce businesses.

It’s a winning formula. The demand for business services already exists among SMBs. And we’re already seeing companies rise to this challenge with the best possible formula for meeting demand like this: marketplaces.

Why AI Agent Marketplaces Will Dominate the Horizontal Angle

There are five reasons AI Agent marketplaces will be a dominant strategy:

• AI isn’t replacing software. It’s turning software into services.

The most successful format for matching labor supply —  or services —  with demand in the last two decades has been the marketplace. SMBs won’t buy from individual providers, they’ll buy from networks of pre-vetted, on-demand AI services they can trust. Like they do from service providers today.

• You need somewhere for all those services to “live.”

If you try to market each such specific AI agent alone – the marketing costs will never justify it. In a marketplace setup, your customer comes to you.

• Standalone marketplaces don’t work for software, but they do work for services.

In software, marketplaces mostly sit on top of existing incumbent platforms (AWS/Azure/MSFT for B2B, Hubspot/Shopify/Wix for SMB, Appstore/Google play).

But for services, which is what AI agents are, standalone marketplaces can work. There are key tactics for building market networks for services, and we have many existing successes like Upwork, Fiverr, and A.Team.

• Incumbent labor marketplaces aren’t likely to adapt to agents fast because they risk undermining their own supply side.

Incumbent marketplaces aren’t likely to build a new system around the AI agent model. They will try to super-power their providers with AI, but have a real conflict launching AI agents who replace their providers. We believe that new standalone marketplaces can emerge using AI agents as a wedge.

• Network effects.

Once a marketplace gains traction, they’re highly defensible (unless you have a large technology shift, like we’re experiencing with AI right now). We cover this here in our network effects bible.

Marketplace building is a science. It’s hard —  but it can be learned. At NFX we have spent years documenting how to do this with human market networks. Now we’re going to port those learnings over to AI, and probably learn many new things along the way.

How AI Agent Marketplaces Will Work

In marketplaces, it’s generally a winner-take-all model. The companies that are already thinking along these lines have a significant head start.

Companies like Enso —  a marketplace for AI agent “freelancers” —  are already demonstrating the potential of AI agent marketplaces.

Right now, Enso actually packages itself as a “vertical AI Agents marketplace for SMBs.” They’ve done this by creating a suite of microservices that any SMB can hire instantly, cheaply, and easily. Enso has put together a market of AI agent LinkedIn writers, SEO experts, Instagram designers and lead finders just to name a few.

All together Enso has launched 300 micro agents, and will have thousands of more soon. These agents have persistent memory, which remember every detail of your business and each interaction you have with them. They start small, execute the outcome well, and learn over time alongside the customer.

And they’re keeping prices low: $49 per month. The fraction of a price of a single service in any other alternative.

The compounding value of agents creates a new form of business asset: one that gets better over time, more valuable with each interaction, and stays affordable enough for SMBs.

Imagine a small local restaurant gaining access to the same quality of marketing expertise as a national chain, or a solo entrepreneur leveraging financial analysis tools that rival those of major corporations. The persistent memory and learning capabilities of these agents mean they become more valuable over time, building an increasingly sophisticated understanding of each business’s unique needs and context.

That’s a unique position, and one that we think is only going to become more valuable in the next ten years.

The Rise AI Agent Marketplaces

The next ten years are going to be about AI agent marketplaces. Especially for SMBs —  who stand to benefit dramatically from this change.

We know the basic rules of how these market networks work. We know that marketplaces can be huge businesses. We even have basic playbooks already written —  but the person who adapts these playbooks to the new world of AI Agents will win.

And like we said, these marketplaces usually have winner-take-all outcomes. That race is already on.