HubSpot Unbound 2026 Recap: 7 AI Updates to Turn On First (And What Each One Costs)

I held about 20 meetings at Unbound 2026, and nearly every company wanted HubSpot's new AI agents. Almost all of them left with homework: get the data (aka HubSpot's Context home) in order first. Here are the seven updates worth your time, what each costs in credits, and the order I'd set them up in.

Watch the recap

This post is the written version of the SmartTake we recorded right after Unbound with Brian Wettach, Partner Development Manager at HubSpot. The full walkthrough is below (about 50 minutes, with live demos). Keep reading for the condensed version, including what each update costs in credits.

HubSpot Unbound 2026 AI Updates: Self-Updating CRM, Agent Hub, Marketing Studio & More. A SmartTake from SmartBug™, Sept. 24, 2026.

What do you need before growth context works?

I spent the week in Boston and held roughly 20 meetings with companies using HubSpot. The conversations followed a pattern. People opened with excitement about agents: research agents, nurture emails that rewrite themselves, a CRM that updates after every call. Within ten minutes, we landed on the same question. Is our data ready for this?

HubSpot's answer, from CEO Yamini Rangan's keynote and the product walkthrough from Chief Product and Technology Officer Duncan Lennox, was a concept called growth context. A prompt gives you an output, and an output only becomes an outcome when the AI knows your business. That means your call transcripts, your emails, your campaign performance, and what worked and what didn't. HubSpot calls itself an agentic customer platform because it holds all of that in one place, so you stop re-explaining your company in a chat window every time.

Here's what I keep telling clients: growth context depends on data readiness. You can build agents without it, and they will return something. You also take on a higher risk of hallucinated output, and every bad run burns credits, because most of the new AI features are credit-based. So we set up the foundation before we build any agents. That's why Context Home leads this list. The other six updates all read from it.

Why does Context Home come before any agent you build?

Context Home lives in the Context area of Agent Hub. It's where you tell HubSpot what the AI should know about your business. You fill in a brand kit (logo, colors, fonts and brand voice), a business profile (tech stack, competitors, stakeholders, products and services), your customers (ICP and personas), your sales methodology, personal info and custom info. Personal info holds your voice and tone guide, which HubSpot uses when you ask it to write emails that sound like you. Custom info takes strategic documents, like how you work with different customer cohorts, with no stated limit on attachments.

HubSpot can fill in some of this by scraping your site. Don't stop there. Most teams already have an ICP and personas, but they live in PowerPoint decks on a shared drive, and someone has to translate them into HubSpot. That translation is tedious, and it decides whether the output sounds like your company.

Knowledge Vaults, which launched in the last six months, group documents that belong together. We built an example finder agent on one. When a salesperson asks what we've done with manufacturing organizations, the agent searches a vault of past projects and pulls the relevant case studies. It searches whatever the vault holds, so the vault needs to be complete.

Think of readiness in two layers. Context Home covers what you document. Underneath it sits the data itself: clean, populated, well-described properties. Every feature below depends on both layers.

Here's the data problem we run into most. A portal has two job title properties, or other duplicate properties, usually because two portals were merged or old data came over from a previous CRM or ERP system without data governance in place. An agent reading both has no way to know which one is correct. Clean that up before you build anything on top of it. Tighter inputs also mean fewer wasted credits, which matters more as these tools get metered.

How does the self-updating CRM work?

Every salesperson has an invisible second job after a call: data entry. It gets done late or it doesn't get done. The self-updating CRM takes that job on. It reads your emails, calls and meetings, whether the transcript comes from the HubSpot notetaker, Teams, Zoom or Google Meet, and recommends property updates. A deal amount discussed on a call or a potential close date can populate automatically.

You choose between approving each update and letting it run on its own. I'd start with approval. Check that what lands in the property matches what was said, then loosen the controls as you build trust. In the data capture settings, you describe what each property should capture, and the clearer that description, the better the AI fills it.

Deal progression builds on that. HubSpot introduced smart deal progression at Spring Spotlight, and this release improves it. After a call, it recommends follow-up tasks, creates a deal record if none exists, and drafts a follow-up email from the transcript. It also gives proactive next-best-action recommendations by comparing what it knows about the deal to your sales methodology, whether that's BANT, MEDDIC or your own training document. You add the methodology in the data capture and deal context sections of Context Home, so the AI knows what your reps should be doing. Deal plans, the shared actions between rep and prospect, come from live call transcripts. We didn't have a live call to demo, so HubSpot is following up with demo videos.

The payoff goes beyond rep productivity. Poor CRM hygiene hurts marketing results, sales forecasts and a manager's ability to coach. It's available with Sales Hub or Service Hub Professional and above, and you opt into each beta.

Why is Breeze Assistant the front door to HubSpot AI?

If you haven't used Breeze Assistant yet, start here. It lives inside your portal and knows which record and tab you're on, so every answer is specific to your business. It also builds things: agents, prompts, workflows, reports, lists, segments, and campaign plans. You can save what works and share it.

Two examples from my week: Customers kept asking how to build specific reports with HubSpot’s AI. I’d walk them through Breeze Assistant, we’d type the prompt, and the report appeared. Then a teammate needed a list of companies matching five filters. Breeze Assistant built the segment in about 10 seconds, and I sent a screenshot of the prompt so they could reuse it. Brian said it has become the first place he goes, like a search bar.

On cost, the assistant itself uses no credits as of today. Once it calls an agent, that agent consumes credits.

What's the difference between Agent Hub and Agent Builder?

Agent Hub is the one place for every agent in your account. It holds the native agents: Prospecting Agent (finds and reaches high-quality leads), Customer Agent (the AI chatbot), and Data Hub (enrichment and data hygiene). It also holds your custom agents, your AI-powered workflows (workflows that include custom agents), and an overview dashboard.

The part I found most useful sits at the bottom. HubSpot looks at how your account is structured and what your users do day to day, then recommends agents and workflows built from that usage. The recommendations come from your own account, so they fit how your team works.

Agent Builder is where you create the custom ones. You describe what you want in conversation, and HubSpot writes the prompt, lists the actions it will take and the Context Home knowledge it will use, and builds the agent. The interface changed in the last week and is fully conversational now. My go-to example is a research agent. Ask it to build a contact research agent that creates a brief before sales calls, and the brief lands on the company timeline. Your rep stops spending the 20 minutes before a call on LinkedIn and news searches. Our team uses this kind of agent internally. The brief can only draw on the company and contact data you have in HubSpot, so complete records matter.

People ask why they wouldn't build this in ChatGPT. Agent Builder has access to your companies, contacts, deals, tickets, and Context Home. It also launches from Breeze Assistant anywhere in the platform. Having AI write the prompt helps too, since tighter prompts mean better output for fewer credits. Native and custom agents both consume credits.

How do Campaign Agent and Marketing Studio change campaign planning?

You reach Campaign Agent through Breeze Assistant or Marketing Studio. Tell Breeze Assistant you need to increase MQLs by 20% next quarter for a specific audience, and it starts asking questions. You can begin from a doc, notes, a file upload, or Insights, which pulls from your AEO performance, past campaigns and segments. One HubSpot colleague of Brian's pasted in the transcript of a brainstorming call and got a campaign out of it.

Campaign Agent then asks which audience segment to target and which campaign to build into. It generates a brief with the goal, narrative, audience, KPIs, and an asset plan based on your portal's performance data. The brief costs 100 credits. Generating the assets costs more.

Marketing Studio is in beta under the Marketing menu. It lists your campaigns and their assets (emails, CTAs, landing pages, social posts), along with performance data, automations, tasks, activity, and budget. The edit canvas looks like a planning tool such as Miro, and it builds the assets as you plan them. You see a pre-show and post-show trade show campaign in one place before anything goes live.

Some context on why this version matters. Marketing Studio 1.0 launched in 2025 and didn't take off, because you had to bring every idea yourself. This version is built around three questions: what do we do next, how do we execute efficiently, and how do we tie it to revenue. Insights draw on your campaign performance and segments, and reporting rolls up once each asset is associated with its campaign. You can use Studio purely as a planning canvas with no AI. Credits come into play when you ask HubSpot to generate assets. If your time is worth more than the credits, build them yourself.

What does Nurture Agent do to an email?

Personalization tokens swap in a first name. Smart content swaps blocks based on a field. Programmable email needs a developer and 47 variations if you have 47 personas. Nurture Agent rewrites the body copy and parts of the subject line and preview text at the moment of send, using what HubSpot knows about the contact. That includes unstructured data as well as properties.

The demo portal belongs to a travel nursing company. The same email goes to nurses and to staffing contacts, and each version is rewritten for its recipient. One subject line reflected an emergency room and trauma specialty, the contact's locations in New York and Chicago, and availability starting in September 2026. Every change comes with a rationale. Brian shared another example: a bank recommending a home equity line of credit based on a customer's equity. We see the same pattern with home builders, where the models a contact is considering can shape the email.

You keep control. Lock the sections you can't let AI touch, like legal disclaimers or tested language. Preview the output for no credit cost, test it on a batch of contacts, then roll out to 25% or 50% of the audience.

Know the limits before you build. Nurture Agent works on automated emails triggered from a workflow. Each email costs 10 credits, about 10 cents, so skip your whole database and use it on your MQLs and SQLs. If the personalization properties are empty, it changes nothing and burns no credits.

This is the clearest case for data readiness: the accuracy and richness of your data decide the output. An enrichment program, whether Data Hub or a third-party tool like Apollo, matters here.

Is Content Agent worth 1,000 credits?

You can reach Content Agent through Breeze Assistant or through the AEO tools. In AEO, open the recommendations section. HubSpot lists topics where publishing content should raise your visibility in AI answers, along with the content type and priority. If you like the topic, click generate blog post. Content Agent writes an outline, takes your feedback, then drafts the post with your brand voice and business context applied. It also creates LinkedIn posts and landing pages, and you can track how often the piece is cited in AI answers.

A long-form piece costs 1,000 credits, or about $10. That sounded high to me at first. Researching and drafting a long-form post takes a writer well over 10 minutes, and at $100 an hour that already costs more than $10. We ran it on key questions about HubSpot integration services, a topic I know well, and the draft covered all the sections that matter.

Content Agent works from your portal's performance data and your business context, which a standalone AI tool never sees. That only pays off if Context Home is filled out, which brings us back to where this post started.

Where should you start?

Here's the order I'd follow:

1. Complete Context Home.
2. Clean up property descriptions and opt into the data capture betas.
3. Use Breeze Assistant daily.
4. Build one custom agent with a clear use case.
5. Turn on Nurture Agent once your enrichment is solid.

Get the context and the data right first, then turn on the features. Which of these are you setting up first, and where is your HubSpot data weakest? Tell me in the comments. We break down every new HubSpot release in our monthly SmartTake. Sign up for the next SmartTake.