Breeze Agents: Your New Digital Teammates | INBOUND 2025

Hello, everyone. It's great to be here. Why did you hire your team? You hired them to do things. You hired marketers to create. You hired salespeople to sell, service people to help your customers with your new digital teammates. AI agents. You need them to be doers. You don't need more helpers. Now, a lot of AI solutions today, they're just helping, assisting your team, creating more content. But they're not doing, they're not taking action. In order to have new digital teammates, they must be doers. And that is what I'm going to introduce you all to today, your new digital teammates. My name is Dylan Sellberg. I've been on the product team at HubSpot for eight years now, and today is uniquely exciting. I've worked on so many awesome products at HubSpot, but this one feels a little different. We are going to change the way our HubSpot customers grow. With AI agents, we are introducing new digital teammates that actually do work inside of HubSpot. Agents that help you service your customers, help you sell, and help you market even better. After watching this talk, I want everybody to set themselves one goal, and that is to have an AI agent live by Friday. Sounds like a tall task, but by the end of this talk, you will see that that is possible. Now, in order to deliver on this promise, we had to look in the mirror and ask ourselves, what is stopping customers from adopting AI agents today? It all starts with a few big questions. All these questions are rooted in the macro, the selfishness that we all carry. What is in it? For me, that is the root of every question about AI. Now, this first question of who is responsible for AI agents really sits at the root of that. Who in your org should be building agents? Who in your org is managing agents? Well, I believe we are about to see the emergence of a new Persona, the AI agent manager. This person is going to sit centrally in an org or embedded within teams, a lot like you see Revops today. And I actually think many of you in this room may even be updating your LinkedIn profiles in the next year or two to be agent managers. I really do. Now, the product we're building today solves for agent managers. Now, the second question a lot of folks ask is, why are AI agents so complicated? Well, they don't have to be. AI agent solutions in the marketplace have been designed to meet complex needs, to take away the intelligence of AI. But what we've designed takes all of the heavy lifting off the agent manager's plate and makes that easy. Now, the Final question we hear all too often is how do agents actually change my business? Now, if an agent is not woven into the fabric of a company, it is not useful. The product I'm going to show today is deeply integrated into the way your company works. It's listening, it's helping, it's, it's triaging and is acting directly inside of the HubSpot CRM. So our goal with what we are launching today is to make sure that you're never asking any of these questions again. Who is building my agents? Why are they complicated? How these actually change my business? We're going to solve those problems and we're going to do that with Breeze Studio. Today, available in public beta, we are launching Breeze Studio. Breeze Studio is your new digital workforce headquarters, the place where you are going to build, manage and deploy AI agents directly within HubSpot. Now, today I'm going to walk through a ton of product. There's so much I cannot fit into a 30 minute talk, but what I'm going to do up front is give you a 90 second demo of all of Breeze Studio. Then what we're going to do is work backwards. We're going to walk through all those component parts, the makeup that make these agents work in detail. All right, so let's start with this 90 second demo. I'm going to talk you all through it. Your journey with AI agents and hiring your new digital teammates is going to start from the new Breeze Marketplace. Here you're going to find over 20 new agents that meet the needs of your business across a variety of use cases. For this example, I'm going to install our Call Recap Agent. Now, the Call Recap Agent has one job. To transform calls into actionable follow up emails and great call notes. Now, before I use it, I'm going to configure it a bit. The job of the agent manager. You'll see it has a detailed prompt already created by our HubSpot team. But I'm going to add some extra instructions. Now, this is easy orchestration. I'm just telling it what to do in natural language. The agent also has two critical pieces of context. These are tools and knowledge. Tools are the agent's hands. Knowledge are the agent's brain, helping it take action and understand your business. Now, again, these agents are complicated. To put them to use, I'll paste in a transcript, enter a company domain name from my CRM and a contact's email address. From there, the agent is getting to work. You'll notice I didn't orchestrate every Single step of this journey, we are leaning into the intelligence of AI. The agent is now reasoning through all the tools it's been given. It's understanding your business's context. It's reading those extra instructions you've given it. It's creating a checklist. It's reviewing its own work. It's understanding all the work you've done in HubSpot. And the result is that call recap. Best of all, this agent took action with its tool. Remember those custom instructions I gave it to include follow ups at the top of its notes? Well, those are there. I also gave it a tool to log its notes directly to the CRM record and it has done that. This agent is actually working for you and your business. Like I said, 90 second demo, a lot of product. Let's back out and talk through what all these components mean make up to be. Now, it all starts with two core concepts. Agents and assistants. Agents, these are your autonomous doers. You give them a goal and they're going to execute tasks for you. Now, Assistants, these are your on demand experts. You can ask them a question and they'll respond with an answer that's rooted directly in your business data. Doers and experts. Simple, powerful idea between everything that makes up Breeze Studio. Now, we've designed Breeze Studio with three core principles in mind to help answer those questions and solve those problems. Our agents must be easy to use. They must be deeply connected to your business and they must drive real business outcomes. And I'm going to walk you through how all this happens, starting with easy to get started. The journey with Breeze Studio is going to start with this Breeze Marketplace. It contains agents across every stage of a journey, regardless of where you're at in your AI readiness. Today, this new marketplace is launching. It is ready for you all to get use and it has 21 ready to work agents included. That library will continue to expand with agents provided by HubSpot, with agents provided by our partners, and eventually with agents provided by you. And these agents are easy to get to work. They are not templates. I. I want to be incredibly clear about that. These agents are ready to work. Once you've found the agent you like, you click install and you get going. That's it. Now, we've wired Breeze Studio directly into your business. Because AI without context is just noise. And because Breeze Studio is CRM native, these agents are always looking at your existing CRM records, your properties and your permissions. All that is happening for you right out of the box. Now let's talk a little bit More about knowledge and tools an agent is only as good as what it knows and what it can do. So we built Breeze Studio using these two new ways for it to gather critical context. Knowledge gives your agent your company's brain. Tools give your agent's hands to take action. By giving agents these knowledge and these tools, you're creating an AI system that truly understands the motion of your business. What makes it unique, why does it work? What guardrails does it need to follow? Diving into knowledge a little bit more, this is your company's brain. You can securely from knowledge vaults, add playbooks, product specs, customer interviews, anything you can fit into a file or already fits in your CRM. This ensures your AI is always working for you, not some generic model. Knowledge vaults currently support your CRM files and hosted content. And when an agent uses a vault, it is grounding its answers in that data to provide you accurate and on brand results. Now, tools are those hands. They can help agents work both inside and and outside of HubSpot. Things like searching the CRM, drafting an email, or even connecting integrated apps. This all happens through tools. We're manifesting tools in a new tools library to give agent managers that context to understand the full catalog of what agents can do for them. And we know tools don't stop with your business. Your context extends well beyond HubSpot. And so we are opening tools to the entire ecosystem. We're releasing a Tools API so that developers and partners can create their own integrated tools, allowing your unique context to travel with your agents. Let's take a look at how we're doing this, starting with our tools library we're launching with over 50 tools, many of them provided by HubSpot do valuable things inside of your CRM, but many of them provided by our early beta partners like happily, Sendoso, Line Pilot and so many more. These give your agents capabilities well outside of HubSpot. This catalog, like our agent catalog, will continue to grow and empower you in new ways to give your agent the ability to take action in and outside of Breeze Studio. When it comes to MCP servers. These enable your business with its own specialized or complex internal systems to connect as tools to our agents. Breeze Studio is built to connect them. We'll provide secure standard ways to register your own MCP servers as tools in some of our early, most powerful tools like Notion, Zapier, Asana. Coming soon Agent AI. These already use MCP servers to give your agents more context. This is yet another way to bring your unique business context directly into HubSpot. Now easy means more than just easy to use, easy to get started. It means easy to customize as well. So I want to walk you through this ease of use journey and how your team's actually going to put agents to work. We knew how important it was to integrate agents wherever your team was working, because our goal is to make it as easy as possible for not only that agent manager, but also the end user. To get value from agents, they shouldn't always need a new tab, a new subscription. They should be able to do that right where they work. So we're launching a new agent inbox. This is the future of how you'll work with your agents inside of HubSpot. Imagine Slack meets Gmail meets Agent Interaction. For every member of your team, this is where you'll take action. Help your agents when they're waiting on you, review their work. It's essentially the home base for for all future agent interaction and that lives directly inside of Breeze Studio. Of course, we're extending well beyond this agent inbox and bringing our agents directly into context. Today we're introducing the ability to embed agents directly in the product. Like you can see here with the company research agent, we're starting with the ability to embed a few select agents on a few select surfaces, like the company research agent in CRM cards, for example. But we'll be expanding those capabilities in the coming months. This makes it easy for anybody on your team to use the intelligence that your agent manager has carefully crafted, but leverage that without leaving their daily workflow. Our agents, they don't need to be told when to work either. They're proactive. The target account, landing page agent, or for example, once it's activated, is always working behind the scenes 24, 7. No lunch, no rest, no breaks. It is always reviewing your CRM data and creating ready to publish landing pages that can help you close accounts more quickly. And it's showing its work directly where your marketing team works in the landing page app. Now, finally, I want to show you how these agents drive real outcomes. This is where transformation comes for business and it's honestly one of my favorite things that we are launching with Breeze Studio and that is agent automation. Starting today, you can trigger any agent based on actual business events. Once you set up an agent to be automated, it's working around the clock. It's listening to the motion of your business and the results, when it's finished, are just there waiting for you. You didn't remember to ask it. You didn't need to remember to ask it. You can run the company research agent every time a form is submitted. You can run the deal loss agent every seven days. You can run the handoff agent anytime a deal is marked as closed one. With Agent Automation, your agents don't remember, don't need you to remember to use them. They proactively get to work. Now, best of all, this experience looks exactly like you're all used to. In building workflows, you can trigger agents off the exact same events in the exact same way. Now, let's wrap up these component parts and then jump into some use cases. We're making Breeze agents easy to use with this marketplace, integrating them throughout HubSpot. We're introducing knowledge and tools to connect the context, the brain and the hands of your business. And we're driving real outcomes with Agent Automation. This stack represents the entire arc of Breeze Studio, and it all begins with the Agent Foundation. This is how it all fits together. Everything is built on this foundation, which includes your core CRM data, a powerful AI reasoning engine, and your integrated apps. This foundation is fused with your business context, bringing together knowledge and tools and your digital teammates. These are your agents and assistants. Your agents as your doers and your assistants as your experts. You'll manage this entire team in a supervised and optimized layer. The marketplace will be where you discover them. You'll use automation to deploy them. Extra instructions and permissions help you govern these agents. And finally, I love the phones go off. That's awesome. You integrate all this power seamlessly where you already work, where your team already works in HubSpot apps in the agent inbox directly from your CRM records. And soon you'll be able to talk to them through Breeze Assistant. Now, I want to show you all of this in action. Let's look at a real company. Well, not quite a real company. A real example of all of our beta customers that we've had in Breeze Studio collected into one. We've created a fictitious company, Seed and Stem. Seed and Stem, they sell business software for landscape professionals. Think HubSpot for green thumbs. Now, Seed and Stem, they were eager to adopt, they were eager to deploy AI, but they had problems. They had problems that were real problems that we saw from the rest of our beta cohort. Their content generated with AI felt generic. Their sales team felt slowed down. AI was reactive. It wasn't proactive in helping them and telling them what they needed before they needed to know, and they had generally more output. But their outcomes weren't really changing. Now, we fixed Seed and Stem's operations with six Agents fundamentally changing their entire business with Breeze Studio. Now, each of these six agents in the context I'm about to provide and share of how they were used and how they were deployed, these are actual results our beta customers have already seen with Breeze Studio. Because with Breeze Studio, Seed and Stem became a breeze power company. Let's start with their marketing team and their personalization agent growing. Their funnel was always an afterthought for this team. The personalization agent. It helps them personalize content at scale to all of their audiences. Now the seed and stem team, they work inside the new personalization workspace. They've installed the personalization agent and as it's installed, it's automatically working behind the scenes to thoroughly analyze their content and their segment data. It's recommending what to personalize. It's generating personalized landing pages based on their target segments and creating landing page variants for each segment. The team shows up to work and they review the agent's work. It gives them justification for why it created a landing page. It makes it easy to edit and publish that landing page directly from the landing pages app. Now the seed and stem team can implement this agent and have more personalized customer experiences at scale. Now, custom assistance helped that marketing team go deeper. The custom assistant that the Seed and stem team built is their Persona assistant. It is a knowledgeable expert of their customer trained using customer transcripts and their icp. Let me show you how easy it is for an agent manager to build something like this. The Seed and STEM team created Handyman Hank. It's a cute name. Now, Handyman Hank. It all begins with a welcome message and then instructions. The message sets expectations for the team. The instructions tell the agent how it should behave. They've given this assistant some knowledge vaults. These are PDFs of customer interviews they performed. And that's really it. They've already built a custom assistant that is ready for their teams wherever they work in HubSpot. Now, moving ahead to the sales team, they were running the copy Paste Olympics. AI was slowing them down, moving text, moving context between systems. To solve this, they deployed the company research agent. Now they learned about the company research agent from the new Breeze Marketplace. From this detailed listing page, they were able to review exactly how the agent worked, what it cost, the tools it provided, even review sample outputs for this agent. And again, it was just a one click install to put it to work. Now, once the agent manager installed that agent, I want to show you how easy it was for them to customize it, starting with extra instructions. Extra instructions are Simple. You type what you want the agent to do and it does it. That's it. Natural language for the company research agent. The seed and stem team wanted one thing different from the default company research. They wanted the agent to always answer if the company had an online booking service available. That data was really important to their sales team and critical for their research. So they just typed in some extra instructions. No prompt chaining, no complex workflows, no trying to guide a superhuman intelligence system to do things that it can do perfectly well on its own. Just tell it what you want from it and the results are there. Now the agent manager does this once and when the team's ready to use it directly from their agent inbox, they get to work. The sales team can use this agent not by memorizing a prompt, but simply typing a domain name. And the agent gets to thinking it will show the sales rep their work and in return it displays this detailed company research report. This helps deliver the sales rep meaningful outcomes in seconds without that complex orchestration that you see in so many agent builders today. Now, if they want to change something about that output, right, every customer has a little bit of a different need. Every sales rep has a different preference. To change that, the sales rep can simply ask directly from the agent inbox. It will incorporate their feedback and get to work again, helping them with their refined results and giving them exactly what they need to move that deal forward. Now the seed and stem team, they needed one more custom assistant for their sales team this time. And this is a sales call assistant. They've got a lot of new reps, they're growing quickly and they have a lot of common questions that need answering. The sales call assistant helps their new reps become top reps more quickly. A sales rep can say, hey, Todd's looking for a discount. The custom assistant is going to answer they instantly if they have a follow up question like what else should I tell them? It's going to reference all that company knowledge and give them detailed plans that the agent manager has uploaded in the knowledge. So now this sales rep has that expert at their side anywhere they're working inside of HubSpot. And now these new reps become top reps more quickly than ever before. Now the deal loss agent truly embodies how agent automation can can work for a business. Seed and stem team use this to solve a strategic challenge and that is learning from their losses. The deal loss agent is scouring the CRM, analyzing deals that have been lost and provides actionable insights to the sales manager without needing to be asked. And they've done this with agent automation directly within Breeze Studio. The agent manager is going to set up this agent automation just like they would a workflow. So this deal loss agent is going to get to work automatically at 5:30 every day. Once the business closed down, they'll look at all the deals that have been loss and when that time comes, the agent gets to work. Nobody needs to tell it to. And when, wherever the sales manager is working, say in the contact record, for example, they'll get a notification, hey, your agent has something valuable for you. They'll bounce over to the agent inbox and have that deal loss report ready waiting for them, analyzing all of their reasons why they've lost deals with intelligent recommendations that are truly rooted in actions and activities that their business had that day. Now this is just one example of agent automation. I think this is going to be the majority of use cases for a lot of early customers who are finding incredible value in agents. In our early beta alone, we've seen 30% of all agent runs have happened using automation. Now, with marketing and sales transformed, the seed and stem team wanted to focus on one final piece of the customer journey, and that is customer success. And this is where Breeze's ability to fuse your CRM data, your company knowledge and external apps that you have, this is where that becomes so critical. Now it all starts with a handoff. The account handoff agent helps sales reps handoff accounts to the CS team with just the click of a button. Let's have a look at how they crafted this agent. Now that Breeze Studio supports integrator tools, the seed and stem team, they have a very valuable integration and that is with Sendoso. They want to send clients gifts each time they get handed off, each time their deal is closed. They've added these two Sendoso tools to their agent. And when the agent gets to work, it's communicating back and forth with Sendoso. So in this example we have the of course the handoff report as guided by the agent manager. But in addition to that, the Sendoso tool was able to recommend the exact gifts to send to that client based on interactions that occurred on that CRM record. With this tool alone, the team who you used to spend 30, 60 minutes manually opening up a new tab, looking through Sendoso, finding a new gift, that's all done for them with the account handoff agent. But their work doesn't stop at the handoff. They wanted to be proactive with their agent use cases, so they enabled the customer health agent, the customer health Agent helps them keep an eye on every account understand opportunities. Now, Seed and Stem, they wanted to teach this agent their account type taxonomy because health means something different to every custom company. What it does is it delivers these CSMs detailed health reports directly to their CRM. So now they're spending more of their energy on relationships and less time digging through CRM data and understanding and making subjective calls about their accounts. Now, the agent manager customized this agent with critical details about their customer life cycle that are unique to the Seed and Stem CS team. They did this with a knowledge vault and they wanted these health reports to be logged directly on CRM records. So they added a tool to create a note with an agent. Now, this agent that came ready to use out of the box has been customized with not only their data but their desired actions. And each time this agent is run, it's going to deliver its results directly to that associated CRM record. So let's recap Seed and stem's transformation a little bit. They easily installed best practice agents and deployed domain expert assistants in seconds. They infused them with their company knowledge, they gave them tools to act, and they automated these agents into the fabric of their business. Now, like I mentioned, Seed and Stem is not a real company, but what they are is is truly representative of our early beta customers. Take Ashby, a modern recruiting platform. Brittany Talty is their head of CS strategy and she is the perfect example of a modern agent manager. She's been using Breeze Studio for a few weeks now. Now her team, they were spending 30 minutes a day on admin after every single customer call. The call recap agent took that 30 minute task and turned it into a 30 second one. That's a 60x productivity gain, giving them hundreds of hours back every single month. Best of all, the Ashby team is thrilled to expand the way that they're using agents in the near future. And it's not just direct customers either. This excitement extends to our partners. Chima is a software engineer at Sendoso. Once he saw his custom tool working in Breeze Studio for the first time, he said the agent was doing some sweet, sweet work and sending some nice gifts. That is the feeling we want every developer, every partner and every customer to have when they're using Breeze Studio. At the end of the day, you have choices when it comes to AI. You can hire developers to tinker with complex frameworks, you can buy disconnected AI tools that don't know anything about your business, or you can use CRM platforms that are incredibly complicated. Breeze is different. What we're going after here is powerful, easy, and lives directly where your business operates. Now, some quick housekeeping to answer or get ahead of. A few questions that I know you all likely have. The first piece on packaging is that today any starter customer can use Breeze Studio. We are rolling it out as we speak. Now, we want you to see the value of Breeze Studio risk free. So before we announce credit pricing for our agents, each agent consumes no credits. There are three exceptions to that. The data agent, the prospecting agent, and the customer agent. But those are all clearly represented on each listing page. Now, when it comes to privacy, your data is backed by HubSpot's standard practices for data protection and privacy. And to better understand those models that we use to power these agents, you can visit BehindHubspotai.com finally, there's so much product left to show here that I wasn't able to cover. We have launched a new permission that allows agent managers to decide who can use and who can do what with each agent. And finally, I know many of you are probably wondering, when can we build custom agents from scratch? Well, you can build custom assistants today. You can clone all of our templates and put them to work for you. And we are actively at work on supporting custom agents and we hope to launch that in the coming months. Overall, with something this new, my recommendation to you all is to just get started, dive right in and start using it today. Because once you do, really only one question remains and that is when will you become a Breeze powered company? Remember that goal I set at the beginning today to to save 30 minutes of your time by building an agent by this Friday. That journey, it can start now. Your new digital teammates, they are ready to clock in and you can learn all about this and more@HubSpot.com agents thank you.