Swift ✉️ Memos

December 30, 2024 | Read Online

2025 is going to be incredibly exciting for AI agents.

I've been deep in the weeds building agent.ai (alongside my day job as CTO of HubSpot), and I'm starting to see some fascinating patterns emerge.

Not just the obvious stuff (yes, AI agents are getting smarter), but also some other far more interesting patterns around how they're going to work together.

So in today's post, I'm going to share 3 predictions for AI agents in 2025:

The Rise of Multi-Agent Networks
Vertical Agents Will Have Their SaaS Moment
Agents Will Start Handling (Some) Digital Life

—Dharmesh

The Rise of Multi-Agent Networks

My first prediction: The future will not be just about singular agents but more about networks/systems of agents where agents can discover and collaborate with other agents.

We've made some amazing progress this past year in creating single-purpose agents. You know, the ones that can handle customer service, do company research, analyze data — that kind of thing.

This has been made possible by some incredible advancements:

Better LLMs with improved reasoning capabilities
More sophisticated structured output
Advanced function/tool calling
And even early multi-modality support (hello, audio and video!)

That's all awesome. I'm here for all of it.

But next year, I think we're going to go much further.

2025 is going to be the year of multi-agent networks.

What I think is coming:

Agents will be able to "declare" their capabilities
They'll be able to specify their inputs and outputs
They will highlight their "experience" (think agent resumes!)

Think about it this way: In the same way that current LLMs support tool calling (whereby a set of tools is made available to the LLM, which it can then use to formulate a more useful response), in the future, agents will be provided access to a network/system of other agents. The agent can then invoke these other agents in order to handle tasks that it can't do itself.

The agent can then invoke these other agents in order to handle some number of tasks that the agent can't do itself.

We're already seeing early signs of this. There are powerful dev frameworks for building these kinds of multi-agent systems (CrewAI comes to mind), and low-code platforms (like agent.ai).

These are early days, and a lot has to happen to fulfill this network of agents vision. But it feels like we're moving this way very, very quickly.

Vertical Agents Will Have Their SaaS Moment

My second prediction: 2025 is going to be the year we see the first billion-dollar verticalized AI agent companies emerge (yes, verticalized AI agents warrant the hype they're currently getting).

Think about what happened with vertical SaaS companies in the last decade. Instead of trying to build one-size-fits-all software, companies focused on specific industries…and absolutely dominated them.

I think the same thing is about to happen with AI agents.

What I think we'll see early on:

AI agents that handle tasks involving collecting and synthesizing information
Agents that can generate digital outputs (documents, emails, reports, etc.)
Agents that know how to use common software tools and platforms
Most importantly, agents that can string these tasks together into useful workflows

The interesting part about this? Any industry that heavily relies on these kinds of tasks — collecting information, creating digital content, using software tools — is going to be transformed.

And that's a lot of industries.

Think about how much of your own work involves these exact things: gathering information, putting it together in a meaningful way, and creating some kind of output.

That's precisely where I think these verticalized AI agents are going to shine in 2025.

Agents Will Start Handling (Some) Digital Life

My third prediction: AI agents are going to start changing how we interact with technology by becoming our digital proxies — understanding our preferences and handling some basic tasks similar to the way we would.

Remember travel agents? Those humans who would understand your preferences and handle all the complex details of planning a trip for you? Then we sort of just moved to do everything ourselves — juggling dozens of websites to book a single trip.

Well, in 2025, I think we're going to see the return of the "travel agent" model… but this time powered by AI agents.

What I think is coming in the near future:

AI agents that understand your preferences and patterns
Agents that can navigate websites and tools on your behalf
Agents that can handle complex, multi-step tasks (without constant supervision)

Example: Instead of you spending hours comparing flight prices, hotel reviews, and trying to coordinate dates, your personal AI agent could handle all of that. Just tell it "Plan me a week-long vacation to Japan in March within this budget" and it takes care of everything.

But it's not just travel. This same model could apply to:

Finding and evaluating software for your business (like choosing a new accounting system)
Managing your personal calendar and scheduling meetings for you
Working with other people's agents to get things done

These agents will improve as they work with you and learn your particulars. They'll learn your preferences, understand your habits, and become more effective over time.

That's it for my three predictions for next year.

I'll sign off this rather lengthy post with one final thought: It's not just about the practical use cases we can already imagine with AI agents — it's about the impractical ones that will suddenly become possible.

Think about all those niche software problems that were never worth solving because the market was too small. AI agents will make it practical to build custom solutions for incredibly specific needs — whether it's for a single company or even a single person.

That's what makes this moment so exciting. We're not just making existing software better, we're making entirely new categories of software possible.

2025 is going to be fun!

—Dharmesh (@dharmesh)