29 min read

A HubSpot Customer Agent + Breeze Conversation

HubSpot has changed. A lot.

If you have been using it for years, you have probably had that moment where someone mentions Breeze, customer agent, Breeze assistant, and “agents”… and you nod like you get it. Then you quietly wonder what any of it actually does.

In this episode, we slow things down and build a simple mental model for HubSpot’s artificial intelligence world, with a special focus on customer agent. This is for the humans who want to use the tools, not get lost in the labels, and definitely not get surprised by credits, settings, or “wait… it can do that?” moments.

 

What You Will Learn

  1. How to separate customer agent (external) from the tools that help you inside HubSpot (internal).

  2. What Breeze means as a bucket, and why it covers more than one thing.

  3. Why HubSpot credits matter, and how to track what is using them.

  4. How customer agent learns from your knowledge base, your website, and even past tickets.

  5. Why “expression” is a big deal for making customer agent sound like your brand, not a robot.

  6. How HubSpot’s growing flexibility is both the blessing and the brain melt.

The Big Moments

Customer agent is outward facing, and that changes everything

One of the cleanest lines we land on is this: customer agent is for your customers. It is outward facing.

That matters because it is not just a tool you play with inside the portal. It is a “front line” helper that can handle chats, emails, and even phone conversations before a human ever gets involved.

The win for the listener is simple. If your team is drowning in repeat questions, customer agent can become the first layer of help, so your humans can focus on the hard stuff.

The knowledge base is no longer just articles, it is your support history

This is where things get fun.

Customer agent can use your written knowledge base, but it can also learn from the treasure chest of past tickets and how your team solved problems before. That gives it a real shot at answering questions like a seasoned support rep, not a search bar.

Then we get into a key idea: when customer agent does not know the answer, it can track those gaps. That helps you see what content you should create next so the system gets smarter over time.

For the listener, this is not about replacing humans. This is about reducing repeat work and turning support history into useful help.

“Expression” is basically custom instructions for your customer agent

George pulls up the area where you can shape identity and behavior.

You can set personality options like friendly, professional, witty, or empathetic. But the bigger unlock is the beta called “expression,” which acts a lot like custom instructions in other artificial intelligence tools.

You can define tone, response style, guardrails, and even scripted patterns. This is how you stop the agent from sounding generic and start making it sound like your team.

What it means for you: if you are going to put an agent in front of customers, you need control. This is that control.

Credits are not a mystery box if you look in the right place

A lot of humans fear turning something on and getting slapped with a surprise bill.

This episode points to a practical fix: go to usage and limits, then look at credits, usage history, and the usage log. You can see what is using credits, how often, and when.

There are also safeguards you can set so things stop when credits run out, or you control when additional credits can be used.

The listener takeaway is peace of mind. You do not have to go in blind.

HubSpot is not “vanilla” anymore, and that is the real reason this takes two episodes

Liz says what a lot of us feel: every time we try to understand something new in HubSpot, it becomes an “oh, okay… what?” moment.

Chad offers a helpful reason. HubSpot used to ship features with minimal functionality, which made them easy to explain but limiting in real life. Now they ship features with a lot more configurability out of the gate, which is better long term, but takes longer to understand.

George adds the bigger truth: HubSpot is all grown up. And we, as users, have higher expectations now because we have experienced more advanced artificial intelligence tools.

So the work is different. You do not just “turn it on.” You bring strategy, goals, systems, and intentional setup.

Practical Next Steps

  1. Write down your biggest five support questions from the last thirty days and decide which ones should never reach a human first.

  2. Open HubSpot and find the credits usage area so you can see what is using credits right now.

  3. If you are testing customer agent, start with one channel first, like chat, before you expand to email or phone.

  4. Review your public website pages and knowledge sources and decide what should and should not be included as agent knowledge.

  5. If you have a knowledge base outside HubSpot, add the external uniform resource locator as a source and confirm it scans the right pages.

  6. Turn on the customer agent “expression” beta if it is available, then write simple tone and guardrail instructions.

  7. Create a clear handoff plan so the agent can escalate to a human when someone asks for one.

Memorable Lines

  • “This is not your grandpa Pappy’s HubSpot.”

  • “Customer agent is outward facing.”

  • “You are basically building a brain with this.”

  • “There is no vanilla HubSpot anymore.”

  • “I am in a sea of confusion right now.”

  • “Oh, okay. What?”

Who This Episode Is For

This episode is for HubSpot users who feel the platform has grown faster than their mental model. If you have heard words like Breeze and customer agent, and you want to use them without pretending you understand them, you are in the right place. It is also for support, service, and operations teams who want to protect human time while still helping customers feel seen and served.

Resources and Links

No links were mentioned in this episode.

Closing

If HubSpot feels bigger than it used to, you are not imagining it.

Take one piece at a time. Start with what your customers ask every week. Build from there. Set the guardrails. Watch the credits. And remember, the goal is not to use everything. The goal is to help humans flourish.

Next step: tell me if you want episode two to focus more on Breeze assistants, Breeze agents, or real use case examples first.

 

TRANSCRIPT

Liz Moorehead: Welcome back everybody.

George B. Thomas: Mm,

Liz Moorehead: I have no

Max Cohen: it back. Run it back.

Chad Hohn: Run it again. Circle it. Let's

Max Cohen: that beautiful bean footage one more time.

George B. Thomas: We're good. No more bean footage.

Max Cohen: Hmm.

Liz Moorehead: So, hi. I don't know why I'm blurry, guys. I, I don't know. Maybe it's a metaphor. Maybe it's because I am in a sea of confusion right now because we're not talking about loop marketing this

George B. Thomas: Mm.

Max Cohen: Do you want it? Would that make you feel better?

George B. Thomas: No,

Liz Moorehead: No,

Chad Hohn: do it.

Liz Moorehead: no.

George B. Thomas: Chad, be quiet. No.

Chad Hohn: Nah, o sh. Okay.

Liz Moorehead: I do have a lot of feelings about Amplify, but we're gonna move on from that. That's for me and my therapist and God. Okay, so we're not talking about Loop marketing because Up Is Down Left is right. Teased Coffee Real Housewives is an actual TV show. I don't know, but I am excited about where we are going today because George, last week when we emerged.

The other side of our massive loop marketing conversation we were talking about where do we wanna go next? You said something that was fascinating to me and then I also deeply related to, you said, I can't tell you the number of times I've heard talk. People talk about customer agent and Breeze as the same thing in HubSpot and they are not, and most folks don't even know there's a Breeze assistant app.

And at first I was nodding along like, yeah, oh my God. Why don't they know that those lessers, those peasants, and then I've gotta be honest, I'm also one of those people, I know they are separate things, but what do they do

George B. Thomas: Yeah.

Liz Moorehead: Unclear. Not sure. And I think that's due in part to the fact that HubSpot has gotten crazy complicated, not just in the past two years, but like if you're anything like the focus on this little podcast shenanigan, right. We've been using HubSpot for over 10 years, and it is. Not the same marketing automation ecosystem that we started with. So if you are today years old, hearing about breeze or customer agent or all of those different things, or you've heard about them and you would just really love nobody to ask you any sort of authoritative questions about them, this conversation is for you.

So George, I wanna kick it over to you. Let's start, let, let's, let's table set a bit here. Okay.

George B. Thomas: Yeah, I, so, yeah, so let me just say a couple things and then you can table set. 'cause I'm sure there was supposed to be a question there, but I started to talk before the question came and, and like, listen, my brain used. You put me in a direction of like, ladies and gentlemen, this is not your grandpa Pappy's HubSpot.

Like I, I look at where we're at. I look at what we can do. I look at all the tools. Like earlier today I was on the customer platform podcast that I do where we talk about HubSpot updates. And I was like, Hey, you remember when like Service Hub was the orange haired stepchild and like now all of a sudden there's all of these like features and things that are happening.

And do you remember when, like we used to have conversations around how, yeah, you know, I'm still gonna use my chat, GPT, my Claude, my X, Y, Z. 'cause it's just not there yet. Do you remember the day when and, and it's like. Man, if you're not, like, I can't imagine. I said this last week, I, I can't imagine even attempting if this today was my HubSpot day one. Like it is mind blowing to me to think about like, um, somebody today purchases HubSpot looks at a marketer that has been there just doing their marketing thing, or sales reps just doing their sales thing. It's like, Hey guys and gals, we bought HubSpot. Here you go. And it's like, oh my God. Like, and yeah.

Max Cohen: Good luck. Have

George B. Thomas: Oh my. Yeah. Yeah. So that's, that's where you put my brain immediately is like this kind of idea of like, it has become, I'll say complex, but not in a bad way. Um, it, it's still very easy to use. It's easy to learn. Um, things are understandable, but there's just so many things. And so how do all of those things fit together and when do you know what?

Liz Moorehead: is technically understandable, but it's also a.

George B. Thomas: Yeah, yeah, yeah. So this gets us to today's conversation, right? Like there's AI in HubSpot and there's customer agent and there's breeze, and what does that mean? Okay. So now I'm sure you had a question you wanted to ask me.

Max Cohen: Can we define those things or try or attempt? I was thinking we

Liz Moorehead: that's,

Max Cohen: do this, George. I'm gonna try to do my best. To define it, and then Chad and Georgie, keep me

George B. Thomas: Okay. Yeah, yeah, yeah, yeah, yeah, yeah.

Max Cohen: so what, which of the ones we wanna define? I'd say Breeze in general.

George B. Thomas: Yep.

Max Cohen: Customer, agent, assistant,

George B. Thomas: That's, that's,

Max Cohen: and what was the other one? That's all three.

George B. Thomas: talk about, uh, let's talk about, we'll, we'll, we'll put it in these three boxes or

Max Cohen: Yeah. Three buckets.

George B. Thomas: Agent

Breeze, general Breeze, custom assistance.

Max Cohen: Okay, cool.

Chad Hohn: About Bri's

George B. Thomas: that's what I'm saying, agents, custom agents or custom. Well, okay. Okay. So I guess that's different.

See? See, here we go. This is the problem. This is the

Liz Moorehead: guys, guys, stop. I want everyone to stop speaking. I literally have, this is the first question with these buckets

George B. Thomas: so hang on. So hang on, hang on, hang on, hang on. So customer agent, breeze, breeze General. Okay, breeze. Um, custom assistance and breeze agents. Those, those are the four boxes. Those are the containers that we're gonna play. See, this is the problem, ladies and gentlemen, those are the containers we're playing in today.

Max Cohen: Okay, cool. I'm gonna do my best to define this, okay.

Chad Hohn: Bucket us up.

Max Cohen: Let's start with just breeze in general, right? Uh, for anyone wondering the way that,

Liz Moorehead: should I

George B. Thomas: No, we need you. We need you, Justin. This is gonna be like herding

Liz Moorehead: max? You literally.

George B. Thomas: cattle, uh, a tornado like

Chad Hohn: having AI hallucinations just like breeze.

George B. Thomas: yeah.

Max Cohen: the question to define this stuff or, no? Am I,

Liz Moorehead: I literally hadn't even given you the, I hadn't

Chad Hohn: asked a question

Max Cohen: Let I thought, sorry, Liz. I thought that was the question

Liz Moorehead: no. George interrupted me before the first question.

George B. Thomas: Yeah, it's

Chad Hohn: Oh, you're just good working ahead on

George B. Thomas: it's like herding tornadoes today. Okay. All right. Um, back to you, Liz.

Liz Moorehead: Okay guys, so what's the simple mental model? People need to keep in mind here about what is what and what does what with your four buckets, right? You got your customer agent, you got your breeze agent, you got your breeze assistant, you got your custom assistant. What the heck is what and why?

George B. Thomas: What a great question. Max, do you wanna try to unpack that?

Chad Hohn: buckets.

Max Cohen: All right. Here's, here's how I understand it. Breeze is, is one, the overarching general brand name of HubSpot's AI at a base level, but also when you start seeing like the enrichment stuff, right? So like. Filling your, you know, filling your, your records with data and like all this other kind of stuff.

That's all Breeze. Customer agent is a very special, uh, AI that is meant to handle things unlike the customer service side, right? So it can do email, it can do a live chat. It can do phone. Now, I believe it can talk to you through your IVR, which is really sweet. Um, but you can train it on all things that a customer service agent would really kind of need to know.

And so it can act as that, like first line of defense, another heat shield, if you will, for your human, your human sales reps to try to get problems solved quickly and get people to the right people at the right time. Um, that's customer agent now. I'm, I'm, I'm, I, I feel like I'm kind of going off an old definition here in terms of the difference between assistance and agents.

All right? Um, I think at the beginning, folks were describing agents as these little AI tools you can build that actually do stuff for you, right? Such as like creating content or doing research or, um, you know, uh, completing some type of task. Then there is assistance. And the way that I was kind of explained assistance at the beginning is that like there are things that like help you do your job, right?

So like there are these little like AI bot that you could invoke through Breeze Chat that will like, give you answers around certain things that it's trained on. Um, however, I've found that you can actually get those assistance to do things.

George B. Thomas: without a doubt.

Max Cohen: So I am, I personally am, am really struggling to draw the clear line between an agent and an

George B. Thomas: maybe we name it different. Maybe it's, maybe they are agents, just to keep this simple, but it's like custom agent versus an agent in the marketplace. Because,

Max Cohen: I'm more

George B. Thomas: because there, because there are, there are agents that you can invoke that have already been built for you. And the reason I'm calling it an assistant is because what we build, and it could be considered maybe an agent, is we build for people, uh, using knowledge vaults and the breeze, uh, assistant, um.

Capability to have custom, and I'll show the screen. I'll show this, but can I just back up for a second and say, max, you did a really good job, but immediately the first words outta your mouth made us, made me realize that we're off the rails already. Because you mentioned a fifth bucket, you mentioned intelligence.

Liz Moorehead: 2.0

George B. Thomas: you mentioned enrichment and intelligence, which I didn't even bring up in the first four buckets. So like, yes Breeze is the general name, but then there are these sub-sectors of that because customer agent could be considered part of breeze. 'cause I'm sure that there's pieces that it's leveraging.

But here's what I don't, I wanna make sure that I get this, uh, out there and then I'm gonna shut up and let Chad go. I wanna make sure that at least one thing we say up front is customer agent is outward facing for the humans that you're helping. And when we talk about breeze ai, whether it's assistance or agents, it's inside of HubSpot helping you, the human, the HubSpot user or users to do things.

So like we have to at least throw that differentiator at customer agent external. Assistant agents, whatever internal, Chad, take a stab at layering on top of or in conjunction with what Max said.

Chad Hohn: Yeah, I think the easiest way to think about just the overall. Breeze bucket as a whole is, it's any of the AI powered features inside of HubSpot, regardless of what they are. So anything that could possibly consume credits either now or in the future. Um, not all Breeze things will consume credits. Like, for example, regular old enrichment just doesn't consume credits anymore, from my understanding.

Um, that used to, but now it's free. Um, you know. So things like that, like basically anything that's AI powered at HubSpot, you could just consider they're putting that under the breeze bucket. That's it. Everything, ai, um, and assistance are chat based, like you pull 'em open in the chat pane or in the mobile app, and assistance are intended to be an ordered, structured prompt.

Driving your conversation and you can add knowledge. But unfortunately, with the regular old assistance, not necessarily extra tools outside of web search at the moment, like you can add web search, you can add files, uh, but if it's just the, I want a specific assistant to do a specific thing, you got a couple prompts, you got some files, it can search the web, it can search your HubSpot records.

Um, that's what it can do, but it, it has basically the regular power that chatting with your regular old Breeze AI chat inside of HubSpot has. So whatever it has access to now, that thing can do quite a bit, believe it or not. So it does have a lot of capabilities, but you can't, at least in the configuration of assistant specify.

I want it to or don't want it to have these tools, for example. But a breeze agent has a lot more prompt configurability and is not intended to be run inside of the chat pane, but is intended to be run inside of the agent execution area. So there's like a specific. Page where the agents execute outside of customer agent, which you kind of like stick it on some channels and it handles those channels, right?

Um, but most of the agents are intended to be run in the agent execution area under the breeze, uh, main nav area. Those, you could give it specific tools. You could give it extra layers of prompting and instructions. And if you take a look at some of the beta ones or the HubSpot ones that exist, those agents have some pretty good example prompts of what you might want to get your breeze.

Agent to do. And it's intended to run a little bit more long form and a little bit less conversational. Give it some specific things. It may ask you for some specific data, um, prior to getting started, depending on how you've configured it. And it's designed to give an ordered output basically. Um, whereas it's a little bit less conversational.

And then, let's see, what was the last bucket there? Um, custom assistance. I think we already went on

George B. Thomas: Yeah, so let me,

Chad Hohn: and then studios, so they have the new dashboards in studio, which are cool.

George B. Thomas: Let me unpack a couple things that you just said, and then Liz, I'm gonna get back to you for another question, but I need to unpack a couple things. So I'm gonna go ahead and share my screen. The first thing that I wanna unpack that everybody would pay attention to, again, if this is your day one, this your day 50, it might be your fifth year, and you're just like, haven't been adopting HubSpot, AI and all that stuff, I want you to realize, one, you do have a place that you can go, whereas usage and limits under your billing area that you can see.

See that where you can purchase HubSpot credits, where you see that you get a monthly allowance of HubSpot credits, your yours may differ than what you see on my screen, uh, as far as the amount that you get per month. But at the end of the day, you're gonna see your usage per month, the credits that have been invoked, and you can also see where.

Where these credits are going. So what is using what, okay, where, where do, where does this get air quotes, money get spent. Um, and then you can see historical usage over time. So again, a little bit of dashboards around credit, and you can even see usage log here. Okay. So I don't want you to think that this is a black box.

I don't think that, I don't want you to think that you're going into this blind, because one of the things that I've heard from people is like. Well, you know, I, I'm, I'm afraid I'm gonna turn something on and then all of a sudden I'm gonna get this massive bill. And like, there are safeguards that you can put in to like, Hey, I'm willing to pay for these additional credits, or when my credits are out, I want it to end.

And, but again, visually being able to see where they're going, what's happening, and if, if there's a monster in the room. You can turn the monster in the room off or, or set some rules around it in, in the future. Second thing I wanna unpack is the marketplace there is, if you go breeze in the main nav and marketplace, there's a marketplace.

These are breeze agents that are already built, right? So like if you're looking for video clip search agent or a brand assistant or an ICP assistant, or, uh, you can just go through this. I'm not gonna read all of these, but there are these. Agents that are prebuilt that you can invoke into your HubSpot portal.

So I, I just want you to know that the beginning places that you may start is, let me go look at the Breeze Agent Marketplace and see if there's something that can save me or a team member time. And then the second place you may go is like, let me see if I can build something that's more specific to us, but then understand.

The credits that may be used if you're, if you go in the customer agent direction or if you go into enrichment direction, or if you go into the, Hey, I created this Custom Breeze assistant and I'm gonna invoke it in a HubSpot workflow to do something type system. So let, let's just, I'll pause there and then Liz, like, what's the, what's the next question?

Where, where you want us to head?

Chad Hohn: Sorry in advance, Liz, but I just wanted to touch on one

George B. Thomas: There we go. Go ahead. Go ahead, Chad.

Liz Moorehead: Chad, it's okay. You were defending me earlier. You get a pass.

Chad Hohn: Okay, thanks. I'm good. I'm glad I'm earned some, earned some, uh, some good, uh, good noodle stars. Okay. So, um, with credits there's not only like. Breeze isn't the only thing that can use credits. Like some things that aren't AI powered, like external data syncs can utilize credits as well. But like George was showing in the dashboard, they give you a summary of every single thing that uses any credits, how much it's using, what days it was using, so you could look at it on like a daily or a weekly basis.

So you really do have a lot of flexibility to understand that. But man is some things that they built in things like Data Studio, being able to sync to an external Google sheet on a periodic time, like without having to create some funky integration. Just your data set, you choose certain columns to go out, that is so helpful depending on what it is you're trying to accomplish.

Um, you know, for certain use cases. So anyway, credits are not only breeze, but there are, but Breeze does, can use credits depending on what it is. So, just wanted to make a point of

George B. Thomas: Good point. Good point.

Liz Moorehead: Is it my turn now?

George B. Thomas: It is.

Chad Hohn: Sorry.

Liz Moorehead: guys. Don't be sorry. This is fascinating. So, um, I gotta be honest, uh, followed a lot of that. Um, also found myself kind of thinking in circles a little bit.

George B. Thomas: Okay.

Liz Moorehead: are the practical use cases like for like if, can we talk through each of those, between five and segment buckets as we have expanded them?

Um, like what is the use case for each, like practical example.

George B. Thomas: Yeah. Yeah. Um, who wants to go first?

Max Cohen: Uh, I can talk about it. Um, what we'll use

George B. Thomas: Let's start with customer agent first.

Max Cohen: Yeah. Yeah. So customer agent, um, so I mean the, the, the big thing with customer agent right, is. Adding this, uh, layer of autonomy, if you will, uh, to the first sort of interactions that someone has, like with your support team, right? So, um, your customer agent can understand a whole bunch of different content you feed into it, such as your knowledge base, uh, a lot of your internal ticketing information.

'cause like, you gotta remember, not only do you have like your written knowledge based articles, it also has a. Treasure trove of information just by looking at your previous tickets and how you've solved problems and things like that. Right. And it gives you sort of this ability to solve problems before they actually get to a human being.

Right. Um, you can almost think about it like before this, like knowledge base kind of solved for it. Right? But like you still had to like. Search through the knowledge base, and a lot of people just wanted to talk to somebody, right? So this serves as like a really cool sort of like middle ground where someone is, is basically at the end of the day, they're interacting with your knowledge base when you really think about it.

But they're doing it in a way that feels like they're communicating with somebody and it might be easier for them to ask questions and then have that thing be able to intelligently find an answer for them, right? Uh, but what's really, really neat about customer agent, is it customer agent, and. We say things like the knowledge base agent and stuff like that.

Those are, those work sort of hand in hand together, but they're kind of like separate. Yeah. Um, so like a really cool thing that the customer agent will do is it'll keep track of like, questions it doesn't know the answer to. Right. So like, when it's finding itself, having to like hand people off to humans or not knowing the answers to questions.

Right. It collaborates with this thing called the knowledge base agent, which actually gives you, uh, you know, a, a sort of like suggestions of knowledge gaps you need to fill with your knowledge base. Right? Um, so like articles that you should be writing, like additional information you should have in there, right?

And what it'll do is it also suggests that content based on answers, it finds looking at like your historic tickets and stuff like that, right? So it's a really cool way to kind of like. Expand and make interactive, um, not only the knowledge based content that you have that's existing, but also just like the past interactions that your support people have had with folks as they've been helping them.

Right. Um, and again, there's. Multiple ways to deploy the customer agent. They can talk to people over live chat. They can talk to people over email, they can talk to people over phone, right? So, uh, again, think of it as like a new sort of like interactive layer and heat shield to your human beings, right? So they can focus on the real, really tough issues, right?

While giving your, you know, your customers a more friendly and interactive way to get the answers that they need, uh, before they have to go to a person, a real life person.

George B. Thomas: Chad,

Max Cohen: Does that help Liz?

George B. Thomas: anything you would add to that?

Chad Hohn: I mean, I felt like that was a pretty good, uh, mechanism, you know, description there. I think one of the interesting things is like too, when you're looking at agent goals. You would think the customer agent could be purely and only support related, but there are, they're starting to like blend the two.

It's just the fact that the customer agent so far is the one that's designed to interact with humans on your behalf at the moment. And they do have like some, oh, what's your lead qualification settings if you want to try and generate leads. Um, you know, so that is a thing that's in there. They're starting in the middle of some sort of blend of moving this into like, you know, I mean, they do have a prospecting agent that exists, right.

You know, that's its own thing. You can apply that to specific, uh, specific, uh, contacts. Like, oh, I wanna, you know, carpet bomb, these 12 people or whatever, you know, or these a hundred people.

George B. Thomas: pretty cool. We might do a, that here's, see, here's the thing. Here's the thing. We might do a whole episode on, like the knowledge base agent. We might do a whole episode on the prospecting agent. 'cause again, right now, all of a sudden we're box six, box seven. Okay? So

Chad Hohn: Yeah. Yeah.

George B. Thomas: ahead.

Chad Hohn: Yeah, I guess the, the main thing I was trying to say is like, just because somebody is in a support queue doesn't mean that they don't have needs for other things that you could possibly have to offer, and that's why they're adding the generate leads section. Like, Hey, if somebody's in support and they meet these extra criteria, maybe it would make sense to offer X, Y, Z if you want the agent to do that on your behalf.

Right. Um, so anyway, that was, that was about the only other thing I would add.

George B. Thomas: a couple things. I'm gonna share my screen and then Liz, we're gonna kick it back to you in a second. So I want everybody to realize, like I'm sharing my screen and so what it looks like, because I said external customer, agent, external, if I start to, um, ask or engage with the customer agent, notice.

I can give it like a couple starter questions that people might actually want to ask to engage with it, but I can come in here. I'm not going to because just a couple different reasons, but I can come in here and I can start to have a conversation with this. Uh. Chat bot, but it's not a chat bot, it's customer agent powered in the background.

And what I want everybody to realize is when you create, uh, the customer agent, you're gonna have this whole area for your customer agent. And really the powerful place is manage and. I want you to understand how far you can go with this based on things that have happened recently in updates. And so first of all, you have your agent identity, right?

And so in my case, I'm showing you sidekick. His personality is casual. You can choose from friendly, professional, witty, empathetic, whatever you wanna do here. Which is great, but there is a beta that everybody needs to know about because this is how you take it to the next level. That HubSpot's calling it expression, but I want you to think about it as custom instructions.

If you're used to chat GPT or clot or projects or anything like that, because you can come in here. And you can be like, Hey, what's the tone? Right? What's the response style? What's the scripted responses? What's the guardrails? And so you can now make this start to act and interact in like crazy ways. And now if you wanna see what this looks like filled out, it might look something like this where it's like.

Sound like a real human who is professional, upbeat, and caring. Uh, you know, you might see something like this where it's like, who am I helping if unclear then? And so like you can, I think it's like 1300 characters or 1500 characters that you can put in these expression. Boxes to really hone, uh, the agent in the way that you want it to interact.

And, uh, always ask a follow-up question. Always ask qualifying questions, things like that. Then you have your knowledge, which is your content sources. So like Max said, the knowledge base. The website, the landing pages, your blogs, and you can upload files if you want additional context that shouldn't live on the site externally, but you wanna add it to the brain of your customer agent.

Not to mention you can add CRM data. So this is where you could literally be like, Hey, I want to go ahead and. Add some properties because, um, this can actually allow them to update properties. So like, hey, if you wanted them to be able to update their phone number or their email or two or three other properties, like you can add these pieces to that as well.

You've got actions which allows you to interact with like. Other, Hey, let them go and actually pay this invoice because it's gonna go through API or whatever, um, channel settings that you have in here. Again, email if you wanna capture, uh, you know, emails for like lead generation. And then where you're gonna deploy it.

So I I, I'm not gonna go through all of this, but the other thing to note is you literally have a performance area where you can look at the last 7, 14, 30, 60, you know, 365 days of how this agent that you're deploying and that you're customizing with these expression statements to really go out there and handle like a lot of the heavy lifting before.

It hands off to a human 'cause at the end of the day, the one thing that I wanna make sure I say here is like, can you use customer agent and never hand off to a human? Sure. Should you have something in place where this does 90, 85% of the heavy lifting and then hands off to a human, when you get somebody like me that it goes, talk to human enter, like, yes, you should.

Chad Hohn: immediately.

Max Cohen: and, and you kind of, you kind of went over this like pretty quickly. I think the other thing too, to make sure everyone's like fully aware that it can do is like, it can make API calls, right? So like, you know, for Yeah. The actions, right? So like, for example, you know. Maybe you are some kind of software company and you have like a password reset process, right?

Like you could trigger it through there. Maybe it needs to go find some data about someone's account. Like you can have it go do that, right? Like so long as you give it the right information and like verify. Yeah. Um,

Chad Hohn: not synced into HubSpot, but you could, you could query out to a different, different database.

Max Cohen: If you've ever chatted with your internet provider and you're like, oh, I want to be, I wanna like try to push a refresh to my router or whatever, like the that, that kind of stuff, like it can do now, which is really cool.

So long as you have the right internal systems for it to talk to. Right. But it can make API requests, which is awesome.

Chad Hohn: Yeah. Yeah. The other thing too is in the knowledge area that I think we glossed over is if your knowledge base isn't on HubSpot, I mean it should be 'cause you should be using Service Hub. 'cause Service Hub is awesome. Them now. It's not just, uh, email inbox like it once used to be. Um, if, if it's in Zendesk or something else, you can actually add external URLs and those URLs associated URLs as knowledge.

So, for example, it'll scan it weekly. Like if your knowledge base is on Zendesk, you give it the homepage and it'll scan up to 5,000 linked pages that are linked to that original page. It'll like go read the knowledge base somewhere else just at a

George B. Thomas: And I wanna make sure that I, I don't want the viewers to just think it should just be knowledge based stuff. Like you can, and by the way, if you're on HubSpot, it'll automatically pull in your pages, pull in your landing pages. Like, 'cause you have to think like every information that you have out there is additional context to the brain that you're building.

'cause you're basically. Building a brain with this. And I want people to know if you, if your website is external, you can give the system a link and it'll bring in your website information, even if it's external outside of HubSpot. So like it's, it really is this idea of collecting all the information.

But then what's real nice is once you've collected all the information into this area, you can go through and you can be like, oh. Oh, this piece of information that we created that shouldn't be in the brain. Like the agent shouldn't talk about this. And so you can say, you know, ki like a removal system once you get it all absorbed into that area.

Um, so it, it gives you the flexibility to really, uh, create the brain you need and train it how to communicate based on the brain that you've now given it. Liz, another question, or should we just move into like breeze? Assistance or like I, by the way, we probably could have done an entire episode on just the customer agent.

Chad Hohn: age. Oh yeah.

Max Cohen: Yeah.

George B. Thomas: Well we wanna do that? Like, do we wanna, with like eight minutes left, do we wanna talk a little bit more about customer agent? 'cause there is something that I could hit on. And then we just talk about Bri's, uh, assistance and custom assistance next week.

Liz Moorehead: that's what I think we're

George B. Thomas: Okay. All right. Let me share my screen again and then guys, we will end up landing the plane, but I wanna make sure that we bring this up 'cause it's a place that not a lot of, um, humans go.

And it's, it really is a place that I would suggest that you go, um, when you're kind of journeying into, especially the customer agent. So you can go into the side panel here, you can go into, um, product updates, and then you can just search by app and you can type in, I'm, I already have it selected, customer agent.

And the reason I'm suggesting this is because one, you can go through and you can look at all the betas. That you can invoke, but also you can take a journey. And whether it's the last two weeks, the last 30 days, or, or the four pages of updates that have happened to customer agent since it was born. Um, you can spend time going through.

Invoking some of these things that we're talking about. So for instance, one of the things you're gonna need to turn on is that expressions beta, right? And so being able to come in here and just, you know, click on here and see like what it is, why it matters, and then if you do or do not need it. You can request the beta or at least then understand how to, um, use actions and API calls or build expressions and things like that.

So definitely look at those updates, um, to the customer agent. If it's your day one of like, we're gonna now use this updates, credits, agent, gentlemen, lady, what is your thoughts?

Liz Moorehead: I have no thoughts.

George B. Thomas: Oh wow. We're in trouble.

Liz Moorehead: I just feel like I've, I'm now kind of glad we're doing a second episode because why is it that every time we talk about something with HubSpot now, it takes us two episodes to get through it. I'm not talking about like our issues as a podcast. It's just like every time I try to wrap my brain around something that HubSpot has rolled out.

I have this one, two punch reaction. Oh, okay. What? Like that is how I felt having so many of these conversations.

George B. Thomas: Yeah. And,

Chad Hohn: might have a little bit of context

George B. Thomas: oh, go Chad.

Chad Hohn: you know, I think the thing that used to happen with HubSpot features is they were rolled out with very limited, minimal functionality. So it was very easy to discuss and to discuss the intended use case, but practically. You would run into the wall of I wanna do this or that custom very, very quickly.

And it wouldn't be useful for anybody who wanted to do anything even remotely outside of the intended happy path, as they call it, in development. Right. Um, the flow that it was designed for now, I mean, it's not that HubSpot is complicated, but it has gotten very. Significantly more configurable, and that's what we're, we're talking about the long tail of all of the configuration options and why those configuration options matter.

But like to put it in context of like, oh, well you could do this and add this expression. Well, why would you want to add the expression? We talk about it for 10 minutes and then there's like six more things that you could configure, but it's really, really awesome in the long run. Because you can just kind of use the boilerplate stuff.

Like what was really cool is when George was showing the expressions area, it says there's like a little AI optimized button. Like all you do is type in like, Hmm, I want it to do this and that and this. And then it puts it in a format that AI's good at reading. You don't need to be a rocket science to figure it out, you know?

Um, but there is a lot of boxes to fill in and it's because they're trying to make things. I, at least I feel they're trying to make things more eff, uh. More flexible out the gate, but that's why it's taken us a lot longer.

George B. Thomas: Well, and I, I think yes to everything that Chad said. Um, but Liz, what I would also say is that the way that I look at this is if I look back at the beginning of this, HubSpot was like an email tool and a dumb blog with some keyword. Like you could see keywords, right?

Liz Moorehead: need to call blocks.

George B. Thomas: No, I know, I know. I'm not calling the blogs dumb, but the tool was just not, the tool was not the greatest tool, but it was amazing.

Max Cohen: They did have smart content though.

George B. Thomas: 2012, I don't know if it had smart content in 2012. I think that came later. Anyway, um, HubSpot is all grown up. And what I will say is we're also, as users, we're kind of spoiled, right? So when you think about this customer agent conversation, it didn't need, it didn't used to be as complex as what we were able to talk about today.

Um, there never used to be expressions, but we got spoiled because we started using Chat t and Claude and we saw the experience that we could create if we could customize the instructions. And so, um, how, how are you gonna. Want us to use a tool that is externally facing and we can't micro fine tooth, how it acts.

And so here's what I'll say. There used to be a day where HubSpot would launch a feature. You would just go use the feature. Now HubSpot launches a feature and you better have strategies and goals and psychology and systems and all the words to understand how do you wanna now set up that tool in a way that it works specifically for your organization, for the humans that you are helping.

Um, because HubSpot has gotten to the point, and by the way, we're talking about AI today, but I would call across the board, there is no vanilla HubSpot anymore.

Chad Hohn: Mm-hmm.

George B. Thomas: used to be able to go and do a HubSpot audit and do a HubSpot audit and do a HubSpot audit. Yo, that ain't the world no more, because it's so configurable and adjustable and you've gotta get in there and get into the trenches.

And so from, from my standpoint, but from the viewer, the listener, there's just a next level that you have to think to get the best out of the tool that they're launching. For you to leverage.

Liz Moorehead: Do you wanna land the plane, George? Or is are we already on the ground? I can't tell.

George B. Thomas: we're close to on the ground. Here's what I'll say. Um, I apologize ladies and gentlemen. Um, I thought that we would get through more than we got through today. So, a, you have a,

Liz Moorehead: Do you listen to.

George B. Thomas: you, you ha I know you have a broad, overarching understanding now of breeze in general. We did do what I'd call a mildly deep dive, or at least giving you things to think about in places to go about customer agent, and I'm super excited to talk next week about custom assistance.

I've had the ability to build multiple for client. And so we'll give you next week some use cases. I'll talk about some things that, uh, we ran into a, a system that I use when I'm actually creating these instructions. Uh, and so we'll go a little bit deeper. And so until then, uh, we'll see you in the next episode.