AI and The Future of Sales, Marketing & CS with HubSpot CEO Yamini Rangan

I think if you're in the tech industry, and I've been in this industry for 25 years, of course this is an unprecedented level of change. But change has been the constant and you just have to thrive in it. I'm like a kid in the candy store right now, looking at all of the new ways in which we can help customers. And so you have to embrace it. And when people around you see that, you're excited about it and you are learning, you are answering questions.

It's really interesting. You said kid in the candy store. Yeah, for me, that's the bar. Like, sometimes people get mad when I say this, especially when I talk to product teams or engineering teams. If they're not saying, this is the kid in the candy store age, you have the wrong team. Yes, they should be excited. They should be going home at night and screwing around and coming up with. They should be doing internal hackathons even though you haven't had one in five years. They should be like, finally some cool stuff they can do. If you're not seeing that, it's brutal. But you got the wrong team.

The last five years were boring.

Boring.

We were doing the same playbook and we were trying to push the same thing. Exactly.

So if you don't feel that it was interesting, even Yamin said it, like, that's a sign. Hey, everybody. This episode is brought to you by adeo, the AI native CRM. Just connect your email and adio instantly builds a powerful CRM with every company contact and interaction you've ever had. Get 15% off your first year. That's 15% off your first year@attio.comsaster that's ado.comsasser. Y thanks for coming back to Sasser.

So good to be back here. Absolutely.

So I couldn't think of anyone actually better to tie to that opener in terms of everything that's changing. I mean, there are. If we look at the public SaaS, companies that are like, classic is B2B. No one's doing much better than HubSpot. But so much change, right?

So much change.

So much change. So I want to get into that. First of all, who in the audience is a HubSpot customer?

Thank you.

Okay, great.

So much.

Most of us. All right, so if Yami is like, if we have time, we'll save a little bit of time for questions because we have so many customers. So if a couple folks want questions, we will save some time. So think about some questions.

Hey, first off, yeah, I have been to Saster for the Last decade, I came in person when I was at Dropbox. I did a video one with you. This is incredible. Congratulations. And the insights that you add to go to market leaders. It's just incredible. Congratulations.

Thank you. Thanks for being part of it. It's good. It's a lot of change. I feel like now, like, there's different eras. There was the pre 2020era.

Yes.

Where it was smaller, but like, we're all doing the same thing, right?

Yes.

And then. Then there was this weird era from 2020 to 2023. Like, first of all, we were all geniuses. Like we talked about before. We were all geniuses in 2020 20. Yes.

I completely agree. I was listening backstage and we thought that we were good, but we were going with the same playbook and getting the excesses of that playbook that is gone.

Gone. Yeah. And then last year. It's good you weren't here last year. Last year was great, like in terms of quality, but it was a low point in terms of this transition vibe. Like, people were still feeling bad for themselves. People were still taught. There's no. Listen, it is not easy at HubSpot today, is it? It's not easy.

Totally not.

But there's no downturn, is there?

No, but we are working harder, harder to get to the same kind of output. I was just agreeing with what you're saying. It is a new playbook. And for a company like HubSpot, we absolutely want to keep all of you existing customers happy. And we are reimagining what go to market looks like in the world of AI. So it's a lot of amps. And that means there's also higher anxiety.

Anxiety, yeah. And how. Since you do, how will you. I see it in what you write. I certainly see in what Dharmesh has done. Right. But how are you helping create urgency at the company? Because I would say of all the public companies, HubSpot seems to be executing the quickest. I'm sure it's not as fast as you want.

It is not fast enough.

We'll talk about it. But it's fairly fast. Yeah. So how have you actually created. Because that's a tough question. How did. How have you helped create the urgency?

It is not fast enough. And my team will hear this from me every single day. Look, I think November 2022 is such a pivotal moment. AI was there before. Generative AI was there before. But November 2022 showed us that you can now interact in a very different way with technology and get more done. And so for us coming into 2023, we just finished the annual planning process. We had the roadmap. We knew exactly what we were doing. We came into January 2023 and we said pivot. That entire roadmap that we just planned for the last three to four months is not going to be what we are building in 2023. And so we pivoted our entire company in at the beginning of January 2023 we started building whole set of new features. The first set of features that Dharmesh helped us launch that are teams launched was March 2023. We launched our first chat spot, now called Copilot in March 2023. Then we launched a number of content features in March 2023. So it was like immediately pivoting. But Jason, I don't think it has been easy. There was just a lot of question. I think you mentioned this 2023 and 24, lot more skeptical voices, a lot more. Yeah, this is going to be like every other big tech fad that was out there. This go away. But I think internally we felt like this is the first time that some immovable metrics in Go to Market can be moved. I've been in Go to Market and I've been in tech for the last 25 years. There are things in sales where what is the percentage of time that your sales team spends in front of customers? I want you to think about this number because this number has been 25% to 35% for more than a couple of decades. We've not been able to move the time we spend in front of customers because we are trying to do all these administrative tasks. We've not been able to expand the account coverage that every single salesperson has. You give your salespeople a hundred accounts, guess what they cover. Actually maybe 40. They cover 40 out of 100. 40% of the accounts get better coverage. So there have been these key metrics in Go to Market that we've not been able to move. That AI is now going to help us move. And so for us it has been just pivoting towards that. And I would say that it comes from the top. One of the reasons, like I do a five minute video for our entire company, we call it a weekly wrap and I send it out every Friday. And I'm demoing what, how I'm using AI and I'm demoing how our teams are using AI. We're talking about where we are getting benefit, where we are seeing real customer value. And so it has to come from the top. So those of you who are leading organizations go to market teams, your founders, you have to lead from the front and you have to believe that there is value for your customers and you have to be able to showcase that value for customers.

Yeah. Let me. I've got a few things. I want to dig in on it, but we were chatting before when we were catching up. Yeah, I don't want. I might have misquoted you. So reframe it. But you said something to me that I've seen, but I still thought it was fairly stunning, which was that on product development side. On coding.

Yes.

Because of you using tools like Cursor and others. Whatever the sweet is, it almost doesn't matter. You now almost have a. You almost have too much product to ship into production. Right. Is it that different than a year ago? I might be slightly misquoting you, but it was dramatically higher output of product.

So I completely agree with what you said just a little bit earlier, which is coding and support. These two use cases have sound product market fit with AI. No question. We're not going back. If you're not doing it, you're already behind. I think coding and support are done right. Like we know that there is product market fit. What I was mentioning is that we started with the first generation of co pilots with coding and it had some level of productivity. There was some value in it. Now we're on to the second generation with things like Cursor with Claude code and obviously mcp. And the combination of all of that is certainly driving the productivity within our engineering organization. We track what percentage of Our teams use AI every single day. Within engineering, it is 95% using one form of agentic code generation every single day. And that means we have to be even more careful about what we ship. It's not just the velocity of shipping something as new features. It is what is the value. Everything in this age. And we'll talk about this, but the business model is changing. It's not like what product you shipped and how much you can get for that product. It is where is there value and how much work is getting done because of AI. And we are becoming even more focused on usage and value that we generate through all of those development efforts.

Yeah. And let's. Let's just talk about support for a minute because not everyone really understands why it's working. Yeah, but people don't understand how good it is now with AI. Typically what I've. Tell me what metrics you've seen. But typically what I've seen is a good AI and Support that it does have to be trained like you usually do have to train it, but you're usually going to be out of the box 10% in CSAT and for depending on how you and maybe deflect 40 to 50% of the issues and maybe resolve another 20 to 30% of routine questions. You still need humans in those cases, right?

You do.

But for an average question, a great, well trained A is better than a human, isn't it?

Here's what we have seen and maybe why is support a great use case? I think because there's a lot more documentation and anywhere there is documentation, anywhere there is data, anywhere there is history, and anywhere there is a known amount of finite work, AI is a great way to solve that. And if you take the tiers of support, tier one support tickets, you should not be using humans at all because you have documentation, you have a support history of how you have solved it and you can get all of this. And so I think internally we started using AI for our tier one support tickets, but we had a counterbalancing measure which is csat. So you can't solve tickets with AI without knowing that you're maintaining or improving your customer satisfaction. So those two metrics go hand in hand for us and internally we've gotten it close to 45% solved with AI. Now last year at Inbound, which is HubSpot's conference, we launched Customer agent and the use case is support tickets. And we have over 2,500 customers now on customer agent. 52% average resolution rate with AI. And we have customers that have between 70 and 80% of their tickets getting resolved with AI. And we are like watching the patterns of these customers really closely. What is the difference between someone who is solving 40% with AI and 80% with AI? It comes down to knowledge and knowledge based articles. And now we launched a second agent which actually identifies that knowledge gap and says from your unstructured data, from your emails, conversations you're having on how you're solving that, we can write a knowledge article and AI writes that knowledge base article and gives it back to the team to look at approve and that is way to improve that overall rate. But 70% to 80% resolution is happening right now. And so it's incredible. In a, in 15 to 18 months we have seen such a huge movement.

And what is, I probably should know this, but what if I'm what how much training? How much time does it take me to get there and have a spot? Like how is it an hour, two hours? Is it doing an extra hour a week like how it's not zero is.

It for it is it's look it's not zero. And it really starts with having well documented knowledge basis and history. The easiest that has taken is get your knowledge base into the agent get all of your support tickets and history into that agent. Sync up all of your website data. You can read now hundreds of web pages in matter of seconds. So sync up your pricing data, all of the website information. Sync up your community answers all of that stuff. Now if you have that readily available it is a matter of few hours to get that in few hours then it is a question of reinforcing are you happy with it? But it's also another point which is where do you want handoff to humans because this is where you started earlier. It's not all AI. AI is augmenting your support teams to drive a higher level of productivity. So where do you hand off and how do you make sure that handoff between agents and humans and sometimes back to agents is really smooth. That's where we are working most of the time on yeah.

And where that this handoff the do you think that ratio will keep going up? Do you think it will can go up to 80%, 90% automated as we get better over the because the rate of change is so fast.

The rate of change is fast. I think that there will still be a portion which is going to be for humans to solve. I don't think there's a lot of hype and there's a lot of you'll get to zero headcount. I don't believe that to be the case. I think there's tremendous value in working with agents reinforcing the learning but also having humans really read the particular case and be involved. So I think there will still be humans augmenting the work that AI is doing and vice versa. So handoffs are going to be really important and that's in general not just in support. I think AI working with humans and handoffs being seamless is going to be important for every function within go to market that is customer facing.

What do you and I want to dig into some HubSpot AI stuff that I have slides next but let me ask you since you're running the leading company that does all of this. I don't know if you heard the story I gave backstage.

Yeah.

Of a leading B2B company where the SDR asked me if I wanted to meet at Saster next week this morning I got an email this morning.

Okay.

Hey Jason. I saw you're going to be at Saster next week. Next week? Okay. This morning. You want to meet next week? This is a leading company. When you hear that in the age of AI, is that going to be gone next year? Is there a role for that human? Because can't AI do better than that human? Do we even need that type of pretty low performing knowledge Professor? Like, how do you react managing an organization that in general and with like how does that, how do you react to that? And someone that's got a sales background that is egregious.

I hope it's not from HubSpot.

Not, I think. But I could tell you three other stories because people ask me, Jason, why does this always happen to you? People ask, but you could do it too. This because I know, yes. I just, look, I just enjoy. I don't enjoy it, but when I get that email, I don't. I like, I have a little fun with it. I make it a case study. I'm not going to put the vendor.

Absolutely.

But I, but it happens every week. I get an email like this, I think. So how's it going to change with AI?

Look, the, the biggest change that's happening is that you have to do more to get your customers and prospects to pay attention to you. There's just an explosion of content, there's an explosion of emails, there's explosion of outreach. And how are you going to cut through that noise? Not by, first of all, if you have like egregiously wrong information like, oh, your conference is next week, then you're out. Like you don't even get 10 seconds in that email and you become a case study for what not to do. But then it becomes a. What does it actually take for a CEO like you to open that email and to engage? It better be exceptionally targeted. It better be personal, not personalized. Personalized. We've been talking about personalization for a very long time, which is you segment your customers, you take your customers and you personalize your data and send it to everybody in that segment. That doesn't cut it anymore. It needs to be personal. It needs to be. Jason, here are the four posts that you did. Here is why I can help you with this single problem that you have. And Let me get 10 minutes with you. Yeah, I sometimes get, and I was telling our sales team yesterday that in the last two weeks I've gotten two videos and they were personal videos and they were. I don't know if it was AI generated or not and it actually didn't matter. But they actually said something very specific about what I talked in the earnings two weeks ago and how they can specifically help. That caught my attention. So I think junk and irrelevant content and both humans and agents not doing their job is not going to cut through. But humans and agents doing their job, understanding your very specific context and sending you a personal message is what will cut through. And we are going through that massive transition. Yeah, it's not volume of content, it is really how personal it is that is going to drive conversion. It's almost like we have to readjust all of our scorecards. It's not. We've gotten used to in the old playbook, increase the volume of leads, increase the amount of pipeline, increase everything. And it's been this focus on volume. It really needs to be focused on conversion because the volume is exploding now it is all about how do you convert and how can you be very relevant, very personal so that you can make that connection. That is what AI is going to help.

Yeah, it's personal biter thing. I'm curious your thought. I've seen this. So sometimes we still get these terror like SaaS for next week. Right. There's another thing I'm sure and I'm curious where you think it'll go or actually what I'm seeing now is great first email, terrible second one. Yeah, the first one's AI. Yeah, I saw on the earnings call subsets doing this. But I saw you have this one issue around personalization of this. We've got a backend solution with Ran and it's great and it's per. And then you're like, great, tell me can I do this? And then I get the worst email after that because all the steps have to be, have to be asked here now to get the attention.

I think the question becomes how is there a handoff? Is the second email written by human? And probably so I think there's a. Throughout the journey, you now have the ability to know what your prospect wants, what your customer needs, what the last conversation was. If you got that first email answered, what was the very specific point that they were interested in? How can you go deeper? I started in the industry in sales more than 20 years ago at this company called Siebel. This was the original CRM company. And before every conversation with the customer, my preparation was like two hours. I would read 10Ks, I would read 10Qs, I would look at transcripts, I would look at anything that was publicly available about that company, their challenges, their strategy. And I would ask very pointed questions and Asking questions was the biggest thing that I learned early in sales. And I think that has become even more important in the age of AI. How can you ask the most relevant questions on what your customer is facing right now so that you can get the next answer? And the next answer. It's like every email, what you're working for is getting the response. Every meeting, what you're working for is getting the next meeting. That's all you have to focus and you have to be just brilliant at it.

Yeah. I worry that as our AIs get better for a while, the handoff for human might get worse. It might get worse because it's a crutch. I'll tell you, the email I got, it was basically an email like this. It was like, Jason, I saw that you just did this interview with Yamini. It's aster. It was terrific. I really liked the point she made about agent to agent conversations. I have a way to create your entire podcast automation for you. I can create the clip, I can redo it, I can do everything your team's doing. I can automate it for you. Here are the three folks. I've done it for you. I can do it. I could even do it for free for you this week. That's a problem I have. It's great.

That is fantastic.

So I responded, I'm like, saster Annual is coming up in a week. I have no time. But if you can do this for real, I'll pay for it right now. This is my response. You won the deal. I get the human response back. Oh, what, what is your podcast and how many do you do? Oh, go to Apple, go to our blog, go to Spotify. We've done 801. Yes. What is your podcast and what do you do? Yeah, that was. You asked the question. That was the human responding to the AI. Yes. The AI was so good.

Yeah.

And I actually said, I'm so busy. Take my money. Yeah, this just. I just. That's a good. That was a good AI, but not a great handoff. Not a great handoff. So I've got a slide here. I know there's a lot here to see, so we don't have to literally go through it, but I wanted to break out two things. Give. Tell me what your. What you think HubSpot will do with AI in whatever time frame. 12 months, 9 months, end of the year.

Yep.

You can talk a little bit about today, but what are you really excited? What are you going to be? We're going to be able to automate in HubSpot, like in the next 12 months. So you're already working on.

Yeah. And maybe a step back took you back to the journey of January 2023. Yeah, I think couple of important decisions that we needed to make back then was there's going to be one product which is an AI first customer platform. And that is an important decision to make because there were a lot of folks building an AI SKU with an AI price and like a separate thing that you would do. And I think for us it is, did anybody buy black and white TV after color TV came out? The answer is no. So you when after AI, there is no product that should not have AI in it. And so I think that was the fundamental way in which we approached our product roadmap. And to simplify this very complex and comprehensive slide, we embed AI in everything that we do. So every product that we have has AI features in it. And we have a copilot, which is a digital assistant, which means any go to market employee should be able to ask questions of the copilot to be able to get answers. And then we have agents that do work for you. Now in the industry, there's just a lot of confusion about what is a copilot, what is an agent? Do you need both? And there's a fight over you don't need copilots, you only need agents. And the simplified version is you actually need both. Copilots help you do work and agents do work for you. And there's a fundamental difference in what is happening with technology. Because forever software and technology, and specifically B2B, we've built it so that. And we've driven innovation and we've poured billions into R and D across all the application companies so that we can help someone sitting in their seat to be able to do more with their time. And that has been the last generation of technology and software. I think what is happening now is that you are not constrained and technology is getting more work done for you when you may not be there. If you have a customer agent and that's answering questions, that website question or the question that comes into you might happen at the middle of the night when nobody is working and that's still answering questions for you. And so there's a fundamental shift that's happening which is how is work getting done? And I give that context because we believe that you need both copilots and agents. And if you look at what you know, we're trying to do is that we launched four agents last year. We added one More agent this year. We're here to prove product market fit. We just want to make sure that every customer that's actually leveraging an agent is getting clear, consistent, repeated value from those agents. So if you're using a customer agent, it should like clockwork, be able to solve your tier one and maybe even tier two tickets. If you're using a prospecting agent, it should look at your entire customer base and be able to tell you what are the best prospects to go after and even help you set up meetings with those prospects. Content agents should look at the distribution of the content that you have and be able to say, here is the best channel to put this content in. Here's how I would cut it for Instagram, here's what I would do for YouTube. It should be able to help you do that work. And that's what we are trying to prove. And what is interesting is that in the world of applications before, it was somewhat predictable. You go in, you point, you click and you can see what the customers are doing. I think in AI it's very new usage patterns. And so all of us in the industry have to be very curious about what those usage patterns are because we just can't go back to what we saw within technology and usage in the last 10 years. It's a new playbook. So we are so obsessed about what are they clicking, which ones have daily repeat usage versus weekly repeat usage versus novelty use. Because what's happening in the industry is there are novel neat AI things. We've all used it in our personal life, we've used it once and it has this beautiful image that it creates and then we never go back to it. That's just neat. But AI has to be necessary. It has to go from just being neat to being necessary to being in the flow of everyday work in order to have the promised transformative benefits that we all think there is. And so that's what we are obsessed.

About is just a couple questions I had. One, you said that. So HubSpot's got I think six core products right now, right? Six as they all have an AI and as you said, AI is core to HubSpot. It's not an upsell, it's not an addition, it's not a 60 bucks a month extra feature. Are those products blending more into one with. Because AI can stretch, it's stretch across them. The difference between sales, support, marketing service. Yeah, if it's one AI I'm communicating with, the products aren't always. They're getting less distinct.

Tell me why you're asking this. This is. I'm very excited that you're getting there. But tell me why you're asking this.

There's two reasons. One, it might be on the next slide, but I'll tell you the real reason. No, I'll go back.

Oh, yeah.

And what I've seen in. I think the bleeding edge here on B2B has been in E commerce.

Yes.

E commerce is because it's super high volume, super transactional. And in E commerce, the lines between sales and support were always blurry.

Yeah.

In B2B, I buy the product often with a human and then support handles the headaches and customer success, smooths over the bumps. In E commerce, when I go to buy some cool shoes or this neat bracelet online, I might want to buy it. I might have a question. Is this in my size? Is this is. Do you have it in gold? Or I might need a return or I might. There's a myriad of things. Or I might come to the site for the bracelet, but I've got to get disaster. Yeah. I might need a coupon. So I go back later. So that's marketing, sales support. And what is what I've seen from the companies I'm close to in E commerce? They're already blended.

Yes.

Klaviyo Gorgias, all these folks in the Shopify ecosystem, they're already merging into one AI and it's almost. It's incredible the pace that it's happening. Classic B2B is. But HubSpot is in this unique position where you have all the apps. So I just think this idea that I talk to this SDR that thinks Sasser is next week and then I have to talk to an AE and then I have to talk to a CSM that doesn't know who I am. And then the marketer's got to get the data and learn like the customer wants. And I'll tell you one less. The customer wants one experience. Yes. And the interesting thing in E commerce, this just literally changed a couple months back. My secret gaming chairs. I'm not a gamer. No. All the gamer. Yeah. Anyone have a secret chair? No. Not enough gamers. All the gamers. You'll watch on YouTube, they'll have these secret chairs. Oh, yeah. It's the. If you're going to sit in your. You're going to stream all day one of.

One of them.

Yeah. We'll get you a seat, we'll get you. And you can brand them up spot. Yeah, brand them. They're pretty cool. So I go to Buy one with my son where I'm at his college, he's a freshman and we're tired and we like bought the wrong chair or something, so I had to switch it. And there's five links, there's chat, our new ChatGPT. There's support, there's sales, there's questions, and there's like literally five chats on this site. I go back, there's one. Yeah, there's one. So I asked that question. This little quote from John Keys, dad, this totally. But not that many in B2B have all the whole stack. Right.

So excited that you're like going here because I think two things are happening. One is that the usage patterns are creating a blur of lines between the different functions in B2B, just like it created in B2C and like other places. And this is exactly right. Like we launched our customer agent and we were specifically thinking about the support use case and we're obsessively watching, like, how are customers using this agent and what is the daily usage patterns? And what is interesting is that they're not only when someone is using a customer support agent, they're also asking about pricing questions. They're also asking about the next event that is getting thrown, the next webinar that you can attend. And so for them it's not, oh, this is a service product and I now can only ask questions about a ticket. It is, I'm your customer and I have questions. That's it. It is not, I'm your question, I'm your customer and I have a service question, a marketing question or sales question. So you literally cannot do that when you put an agent out there. And this is one of the realizations that we had. The lines are blurring and at the end of the day, you serve your customer. Your customer does not need to know which work they're dealing with. And we've been saying this for a while in our industry, it is now possible to do. I do think the lines are blurring. And in order to do that, the level of documentation, the level of history, the full context of the customer journey needs to be there. And we do think that's one of the reasons why platforms will continue to win. Point agents will not win in this whole race that we are in. That's one part of it. There's another subtle thing that is happening, which is we have in B2B software always focused on providing technology for specialists. BDR has a tool, a pre salesperson has a tool, a solution consultant has a tool, a salesperson has a tool, a content manager has a tool, a video editor has a tool. Like there is. We've really provided tools for specialists. And what AI is going to help us do is to enable generalists in marketing. You might not be.