HubSpot AI Explained: Breeze, Agents, Agent Hub & How to Get Started

AI is everywhere in HubSpot right now.

And that's exciting—but it can also be a little overwhelming.

If you've been hearing about Breeze, AI agents, Agent Hub, AI credits, prospecting agents, and AI-powered workflows, you might be wondering where you're actually supposed to start.

That's exactly what we wanted to address.

In this Q&A, we answer five practical questions about using AI in HubSpot—and, more importantly, how to use it without turning AI into another shiny object your team plays with for a week and then forgets about.

What Is Breeze in HubSpot?

Let's start with the simplest piece of the puzzle: Breeze.

HubSpot has grouped its AI capabilities under the Breeze brand. One of the easiest ways to think about Breeze is as the AI assistant that lives inside your CRM.

You can use it to ask questions, retrieve information, and help with everyday tasks.

For example, you might ask it about a contact, how to accomplish something inside HubSpot, or information stored in your CRM.

Think of it as a conversational layer sitting on top of your HubSpot account.

It's useful—but it's only one piece of HubSpot's broader AI strategy.

What Are HubSpot AI Agents?

AI agents are a different concept.

Instead of simply answering a question when you ask one, an agent can actually perform tasks for you.

That could mean executing a task once or running an ongoing process based on certain conditions.

Some examples include:

• Prospecting agents
• Customer handoff agents
• Call recap agents
• Data-related agents
• Lead scoring agents
• Closed-lost analysis

The important distinction is that an AI assistant helps you do something, while an agent can potentially be configured to do something on your behalf.

That's a pretty significant shift.

What Is Agent Hub?

Agent Hub takes things another step further.

Historically, HubSpot users could access agents built by HubSpot and other providers through the marketplace.

With Agent Hub, organizations can also start building their own custom agents and agentic workflows.

That opens up a much bigger strategic question:

What repetitive work inside your organization could an AI agent actually take off someone's plate?

That's where the technology starts becoming particularly interesting.

But there's an important caveat.

Don't start by trying to automate everything.

Start Small With One AI Use Case

One of the biggest mistakes companies can make with AI is trying to implement too much too quickly.

Instead, start with one very specific use case.

For example, imagine you have an agent that runs after a sales call and creates a useful recap.

Or maybe an agent monitors closed-lost deals and analyzes the CRM activity, meeting transcripts, and email history to identify potential reasons the deal was lost.

Those are relatively contained use cases.

They have:

1. A clear triggering event
2. A defined task
3. A useful output

That's exactly the type of process you should look for when getting started.

Once your team sees the value of one successful AI implementation, it becomes much easier to identify the next opportunity.

How Do You Control HubSpot AI Credit Usage?

This is another important consideration as organizations start experimenting with AI agents.

Some custom agents consume credits, so you need to understand how much you're allowing your organization to use.

HubSpot provides controls in billing settings that let you set usage limits.

That's important because you don't want an experimental agent running unexpectedly and generating a large bill.

The strategic advice is just as important as the technical setting:

Start small.

Run one agent.

See how frequently it executes.

See how much value it generates.

And monitor how much of your available credit allocation it consumes.

Then you can make a more informed decision about expanding your AI usage.

Can You Trust HubSpot AI to Write Emails?

This is where things get a little more nuanced.

AI can absolutely help you draft emails.

But that doesn't necessarily mean you should immediately let it send those emails automatically.

A better progression is:

Draft → Review → Refine → Automate

For example, HubSpot's prospecting capabilities can identify prospects who may be ready for another follow-up and generate a suggested email.

That's incredibly useful.

But initially, you may want a salesperson to review that draft before it goes anywhere.

Over time, as you understand how the system performs and improve its instructions, you can decide whether certain pieces of the process are safe to automate further.

AI Is Only as Good as Its Inputs

One of the most important principles in AI implementation is that the output depends heavily on the inputs, instructions, tools, and knowledge available to the system.

If you haven't given your AI enough information about:

• Your company
• Your brand
• Your ICP
• Your products and services
• Your tone of voice
• Your processes
• Your dos and don'ts

… so the result is probably going to be generic.

And nobody needs more generic AI-generated content.

The goal isn't simply to turn AI on.

The goal is to teach the AI how your organization works.

Should You Let AI Build HubSpot Workflows?

HubSpot AI can help create workflows from prompts.

That's useful—but it's still not a replacement for a human who understands your CRM.

Think of AI-generated workflows as a starting point.

You still need to:

• Review the logic
• Customize the actions
• Check enrollment criteria
• Test the workflow
• Make sure the automation won't create unintended consequences

In other words, don't go from zero to 100.

Let AI accelerate the work while you maintain the strategic layer.

Is HubSpot AI Worth It for Small Teams?

Absolutely.

In fact, smaller teams may have some of the most to gain from AI.

Imagine a four-person sales team implementing a handful of useful AI agents.

You can test them quickly.

You have fewer people to train.

You can identify problems quickly.

And when something works, you don't have to roll it out across hundreds of employees.

That makes smaller organizations an excellent environment for experimenting with practical AI use cases.

The Best Way to Start Using AI in HubSpot

If there's one takeaway from all five questions, it's this:

Start small.

Don't begin by asking:

"How can we use AI everywhere?"

Instead, ask:

"What's one repetitive process that happens regularly, takes meaningful time, and has a clearly defined output?"

That's where you should start.

Build one use case.

Test it.

Measure it.

Refine it.

Then find the next one.

AI in HubSpot isn't really about checking a box that says your company is "using AI."

It's about building systems that make your people more effective.

And that's where the technology gets genuinely interesting.