Top HubSpot AI Features For Marketers To Use In 2026

We’re sure you’ve noticed how HubSpot has been busy embedding AI into every part of its platform.

There are now more than 80 embedded AI features by the company’s own count, a set of prebuilt agents, a builder for making your own, and a separate product for tracking your AI search performance.

There’s a lot to like in here, and some genuine game-changers. The trick is working out which offerings suit your team, what each one costs to run, and how much setup you need before you get paid back.

Read on as we cover exactly HubSpot’s AI features, along with use cases and strategies for getting the most out of them.

1. Breeze Assistant (formerly Breeze Copilot)

Breeze Assistant is the chat panel that sits alongside whatever you’re doing in HubSpot. It reads your CRM and knows your role, allowing it to personalize its responses.

You can ask questions such as:

Summarize a contact before a call
Draft a follow-up in your brand voice
Pull together everything that happened on an account since any month
Explain why a campaign underperformed, and much more

It comes with every Core Seat and is also available on free plans, subject to usage limits.

Breeze and Loop Marketing,

At the Spring 2026 Spotlight in April, HubSpot retrained it on Loop Marketing, the framework it launched at INBOUND 2025 to replace the funnel.

For example, if you ask for campaign help, Loop steers you toward AI-driven visibility and personalization rather than classic top-of-funnel content, naming the HubSpot tool that handles each stage.

Its main weakness is the quality of context, rather than access. Breeze Assistant can pull from CRM records, connected apps, uploaded files, saved memories, and HubSpot’s own marketing guidance. You need to provide good detail across those sources for solid results.

2. Agent Hub and the Prebuilt Agents (formerly Breeze Agents)

Agent Hub went into public beta in July 2026, first findable under More, then Agents.

It’s a console rather than a product, and it was unified because HubSpot’s agents were scattered across five parts of the platform. It’s Professional and Enterprise only, and offers functionality across Marketing, Sales, Service, Content, and Data Hubs, plus Smart CRM.

HubSpot’s agent army

The lineup covers most of the repetitive work in a go-to-market team, though the quality varies enormously between them. Some key agents include:

Customer agent: Resolves support conversations across chat, email, and messaging using your knowledge base and CRM history. HubSpot reports it resolves 65% of conversations on average, with the strongest teams reaching 90%.
Prospecting agent: Watches accounts for buying signals, sources contacts, and drafts outreach.
Content agent: Drafts blog posts, landing pages, and case studies from a brief and can directly turn AEO recommendations into publishable content.
Knowledge base agent: Reads support tickets, finds the questions your documentation doesn’t answer, and drafts the missing articles.

Jeff Pedowitz, whose company has run more than 400 HubSpot portal audits, identified that “The marketing is ahead of the reality in some areas and behind it in others.” He found the prospecting agent cuts account research time by 60-70%, while the content agent cuts production time by 50-60%, though it requires proper human editing. Since April, both flagship agents bill on outcomes rather than usage. “Too often, that means paying for potential rather than performance,” said Jon Dick, HubSpot’s Chief Customer Officer. “Outcome-based pricing removes that risk. You pay when it works, full stop.” However, that means you need to understand the outcomes to budget properly.

Breeze Assistant opens in a panel beside whatever you’re working on, so a question about improving a page is already scoped to the page in front of you.

3. Agent Builder (formerly Breeze Studio)

As the name suggests, Agent Builder lets you build your own custom agents. It’s in public beta and included with Professional and Enterprise. You describe what you want in plain language, and it assembles the agent from your CRM data.

Agent Builder covers most of what a marketing operations team needs, but it stops short of a development environment. You can do the following:

Choose what starts it and what it covers: Pick the trigger, decide which contacts or companies it should work through, and say where the answers get saved.
Choose its freedom: Approve every action yourself while you tweak, then let it work autonomously once you’re happy with the outputs.

Agents built with HubSpot aren’t limited to HubSpot either. An agent can browse the web, read and write CRM records, use uploaded documents and knowledge vaults, and connect to external systems through MCP. HubSpot currently supports connections with Notion, Jira, Confluence, Asana, Zapier, G2, and Linear. So there’s plenty of freedom for more advanced implementations.

Matthew Alex Ruxton, Senior Manager of RevOps at Ignite Reading, described what the custom agents did for his team:

“It used to take 15 to 20 minutes to find the information we needed and copy it into HubSpot. Now it takes seconds with a custom agent in Agent Builder. Based on how often we completed this action last year, we anticipate getting roughly 350 hours back from a single automation.”

That’s a sensible first build – one repetitive lookup that eats 15 minutes, hundreds of times a year. Start simple and expand functionality as you gain confidence rather than going from 0 to 100 with your agent’s complexity.

4. Data Agent

Data agent is available from Starter upwards, and it’s cheap to run.

A typical use is researching questions using your CRM, call recordings, documents, and the web, then writing the findings into a smart property. HubSpot’s own examples give a sense of the range of questions it can respond to:

Qualification questions: Does this company need a sales team, and is there anything suggesting they’re in a buying cycle now?
Structural questions: Does this company have a European headquarters, and what products does it sell?
Relationship questions: Has anyone on my team spoken to this person about a given product yet?
Signal questions: Has their CEO appeared in the press lately, and which customers do they name as happy in their case studies?

Be aware that, currently, the data agent bills per record, not per question. Run the same question across 5,000 companies, and that’s 5,000 separate responses, consuming 50,000 credits.

At the standard additional-credit rate, that’s $500 of credit usage before accounting for whatever credits your subscription already includes!

You can see the console view for the data agent on the right, with a walkthrough to get to grips with it.

5. HubSpot AEO

Answer engine optimization (AEO) is the newest tool here and the easiest to justify. As you know, buyers are using ChatGPT, Gemini, and Perplexity before they visit anyone’s website, and those tools answer with trackable names.

HubSpot AEO launched in April, built on XFunnel. It does four things:

A visibility score showing how often you appear across the three engines.
Prompt tracking, where a prompt is one of the questions your buyers type into an answer engine.
Citation analysis, which shows the domains and pages the engines pull from when they answer for you and your competitors.
Prioritized recommendations telling you what to publish or fix next. Most teams underutilize sentiment scores, ranging from -100 to +100, because a model can mention you and still steer the buyer elsewhere.

AEO currently costs $50 per month standalone or $45 per month when paid annually, with 25 tracked prompts and no HubSpot subscription required – so you can use it as a standalone product. Marketing Hub Professional and Enterprise include AEO with more prompts. Before you start using it, run the free AEO Grader. It runs a set of test queries and returns a score out of 100 across sentiment, presence, brand recognition, share of voice, and market competition in about five minutes.

The same brand in the same week, scoring 46% on Gemini and 3.6% on ChatGPT.

What HubSpot AI Costs to Run

Some of HubSpot’s AI runs on credits, and you can find the rates in the product catalog.

Marketing, Sales, Service, Content, and Revenue Hub include 500 credits a month on Starter, 3,000 on Professional, and 5,000 on Enterprise. They reset monthly, don’t roll over, and aren’t additive if you operate several products.

Data Hub and the Customer Platform are more generous at the higher tiers, including 5,000 credits on Professional and 10,000 on Enterprise. Additional usage is $0.01 per credit, or $10 per thousand.

The spread between actions is wide and hard to get a handle on, currently looking like:

Data agent: 10 credits for one prompt against one record.
Customer agent: 50 credits to resolve one conversation on a text channel.
Prospecting agent: 100 credits to recommend outreach for one lead.
Content agent: 1,000 credits per piece, which is $10 an article and a hundred times what a data agent prompt costs.
Custom agents: 1 credit per action unit, so cost depends entirely on how many steps you give the agent.
Breeze actions in workflows: 10 credits per action, so an AI step inside an automation is metered even though the workflow itself isn’t.

All costs are worth modeling before you commit. For example, with the content agent, a team publishing three pieces a week can spend some $120 a month on drafts alone.

Overall, you’re best off choosing the best features and going strong on configuration and setup, and auditing your results. Spray and pray is costly and unnecessary.

Where to Start With HubSpot AI

Start with whatever your plan already includes and costs the least to run, which for most teams means the data agent.

Start with one high-value property, such as whether an account is showing hiring or funding signals. Run it on a small sample and check the results manually before applying it to thousands of records.

If your plan includes the agent builder, resist the urge to build something ambitious first. One repetitive lookup a person currently does by hand makes a better first test than a multi-step agent.

Run the free AEO Grader, regardless of which plan you’re on. It takes a few minutes and costs nothing. You’ll get a good idea of AEO visibility, but it is not the be-all and end-all, as it’s not 100% accurate and doesn’t query the internet (meaning only training data) across Gemini or ChatGPT yet.

As always, if your records aren’t in reasonable shape, fix that first before doing anything, as the old AI adage stands: garbage in, garbage out!

Want to go deeper on building personalization that moves revenue? Register for our upcoming Virtual AI for Marketers Summit!

Here are some similar articles you might enjoy:

Author

Sam Jeans

I’m a writer and marketer specializing in AI, search, and emerging technologies, with particular expertise in SEO, AIO, and LLMO. My work explores how AI is changing the way brands create content, reach audiences, and compete for visibility across both traditional search and AI-driven platforms. I’ve also helped develop AI-led ranking techniques used by organizations including Apple, Waymo, and Forrester.
My broader writing and marketing experience spans SaaS, enterprise technology, machine learning, and Web3, with published work for organizations including Gartner, Panasonic, Epicor, and The Independent. I bring a mix of technical understanding, marketing strategy, and editorial experience to my writing, with a focus on making complex AI developments practical and useful for marketers.