HubSpot Spotlight: Dharmesh Shah | INBOUND 2023

Greetings, fellow humans. And a shout out to all the AI bots out there. I know you're watching too. I'm Dharmesh. I'm one of those co founders at HubSpot that consumes everything on the Internet about Generative AI and sends it all to Andy. By the way, one more round of applause for Andy. I'm so proud of the team and what they've delivered. It's been amazing. Now, originally I had thought to make this presentation mostly about AI. You'll be pleased to know I changed my mind. It's now all about AI. Not just because AI is the topic of the day, which it is, but because it's also the opportunity of a lifetime. And not some random lifetime. I'm talking about you and your lifetime. Yamini talked about how we can leverage Generative AI to connect better with customers. Andy just showed you a bunch of amazing things the HubSpot team's been working on to help you deliver on those ideas. I'd like to talk about love. I have an unhidden agenda. My mission is to have you fall in love not with AI, but with a future version of yourself that's AI powered. That's why we're here. So I'm going to first start off with the story of how I fell in love with Generative AI. I'm going to try and convince you as to why you should be enamored too. We'll talk about where the relationship is headed and then what I think you should do, otherwise known as unsolicited relationship advice. So let's start with the story of how I fell in love with generative AI. The story goes way back to 2 BC. And by BC I mean before ChatGPT. It's the summer of 2020 and I get this weird message. Brian, my friend and co founder, asks me if I want to join him on a call. And my immediate reaction is, wait, what? You see, I'm an introvert that dislikes phone calls. Otherwise known as an introvert. My friends and family have learned that the best time to call me is email. And you know what's weirder than that message is the fact that I said yes voluntarily. Didn't even have to be bribed with a donut. By the way, the one thing better than a chart that looks like a donut is a donut that looks like a donut. I'm just saying, I made that up on the walk up here. Sorry I tried out on you. So why did I agree? Because this call was with an entrepreneur I had known and admired for a long time. He was working on a cool AI startup and I had been obsessed with AI machine learning for a long time. Now I didn't know this at the time, but that startup would end up having a major impact on my life on HubSpot and as it turns out, the entire human species. The startup was OpenAI and that entrepreneur was Sam Altman. I am a big fan of OpenAI and full disclosure, I'm a small investor in OpenAI. It was a mind expanding call, but there was one comment from Sam that really stuck with us. He said AI will progress gradually for a while and then at some point going to hit a breakthrough and things will start to move really, really fast. He was totally right. Soon after our call I reached out to Sam. I emailed him like a proper introvert and I asked him if I could get access to the GPT API. I got a lightning fast response back. The man is a machine. No, seriously, he actually might be part machine. So now I have access to a GPT API. So I do what any reasonable AI obsessed developer would do. I started coding. I built a simple chat app and have my first conversation with generative AI. I am in love. I'm just a boy sitting in front of a generative pre trained transformer asking it to love him. I spent many nights over the weeks and months thinking about GPT. And then in November of 2022 the little GPT I fell in love with became a big star known as ChatGPT. ChatGPT is amazing. I can ask it to draft an article, craft a marketing strategy, help me write a song about AI. ChatGPT is smart, powerful and makes me a better human. Just like my wife Kirsten, by the way. She has endured multiple practice runs and has heard all five of the dad jokes in this presentation multiple times and laughed at all of them. If that's not true love, I don't know what it is. All right, so back to GPT. I discovered something even more amazing about it. Yes, it can create things for human consumption, but it's also shockingly good at creating things for computers, like code. And this was a very big deal because although most of the current excitement around generative AI is about creating things for humans, like text or images or video, a lot of the future opportunity is going to be because it can generate code, because that opens up amazing useful opportunities. All right, so let's look at an example. Here's how pretty much every software application works. There's something you're trying to get done like building a rapport, doing something. And to get what you want, you have to go through a very specific series of steps. You point and click and drag and drop and touch and swipe and then you make a sacrificial offering to the software gods. And then maybe you get the thing you were trying to do. But now with the power of generative AI, we can build natural language interfaces to software. So instead of going through a bunch of these little steps, you can just ask for what you want, type it into a prompt and boom, it's done. I was super excited about this idea of building a natural language chat UX for HubSpot. So I did what any reasonable AI obsessed developer would do. I started coding. I called the app Chat spot. It's like ChatGPT and the HubSpot CRM had a baby. Now, have you ever worked on something all day and then late into the night and you wonder where the sun went and then it dawns on you? That was me for many nights. And we launched chat spot on March 6, exactly six months ago. And since then we've had over 80,000 people chat with it. Now you can do all sorts of things like company research, keyword analysis, access to HubSpot CRM and a bunch of other things. I encourage you to check it out at Chatspot AI. There's a session tomorrow morning at 9:30, HubSpot Hero Stage, where the team, there's an actual full team now working on it. It's an exciting demo. I've seen it. It's awesome. By the way, you don't have to be a HubSpot customer to use Chat Spot AI. It's free. You don't have to be a HubSpot user. You don't even have to laugh at my dad jokes. It's worth it. All right, so that's the story of how I fell in love with generative AI. Now let's talk about why you, the good people of inbound, should be enamored with it too. So let's first compare generative AI to some of our previous relationships with technology. When we first got computers, they had an additive impact to our lives. It was human plus computer and computers provided computation at scale. Then the Internet came along and it was awesome. And it wasn't just about the cat photos. Okay, it was a lot about the cat photos. But the Internet multiplied our capabilities. It was human times the Internet. The Internet provided information at scale. And now we have generative AI, which has an exponential impact. It's humans to the power of AI. This is not a once in a lifetime thing. This is once in many lifetimes. Because generative AI provides cognition at scale. In a way, it does this like a toddler might. A toddler watches us, learns from us, mimics us. The mocking comes a little bit later. What's amazing is that GPT4 can now ace standardized tests. Most toddlers can't score in the top 10% on the bar exam. Most lawyers can't score in the top 10%. Well, I guess technically only 10% can anyway. All right, so you might be thinking, good for GPT. I'm sure its parents are very proud. But what does that have to do with me? What does that have to do with my company? So let's look at some numbers. McKinsey put a report out and they project that generative AI will add over $2.6 trillion of value annually. That's trillion with a T. There are other letters in the word, but it's the T that makes the number impressive. And of the 63 use cases they identified, three of the top five are sales, marketing, customer operations. So generative AI can help us market better, sell better, service customers better. But it's not all donuts and dandelions. There are some red flags in the relationship with generative AI and we should walk into the relationship with our eyes wide open. Generative AI has sometimes has what are called hallucinations. It makes things up just like some humans, and it says it with 100% confidence. Just like some humans, it just doesn't know what it doesn't know. Just like some humans. Generative AI also has some other hazards. It can generate responses that are inappropriate, biased, or sometimes even harmful. Now, the good news is that the AI providers have been working really hard on this problem and have made measurable progress. Still, my advice is to keep humans in the loop. When a relationship is moving really, really fast, it's important to invest the time to build trust. And AI is moving really, really fast. So let's look at where this relationship with AI is headed. Turns out I have an inside advantage when it comes to figuring out where AI is headed. No, I don't have Sam Altman on speed dial. I don't have anyone on speed dial. What I do have is a 12 year old named Sohan. He lives inside my house. He hates changing out of his pajamas and he loves all things tech, including AI. I don't know where he gets it. Three years ago, three years ago I showed him GPT when he was in his early nines. Now he Uses GPT in more creative ways than than I do. Recently he showed me a role playing game he built entirely inside ChatGPT. His prompt was hundreds of words long. So what he was doing is basically giving the computer instructions in what I think will be the most popular programming language on the planet. Not Python English. After playing his game, I'm like, this is really cool little dude. And he's like, no cap bruh. The parents in the room will identify. It's like, okay. But Solhan doesn't just use generative AI. Solhan is a member of Generation AI, the entire generation of people that's growing up natively with this technology, just like many of you grew up with the Internet. Not me though. I grew up having to look up things in a book like a caveman. Anyway, to see where AI is going, all I have to do is see what frustrates Sohan, because whatever he's frustrated by, it's going to get fixed not in decades or even years, but in months. So let's look at some of the frustrations. Frustration 1 Right now, large language models are a static snapshot based on when the training data was taken. That's why ChatGPT and GPT4 doesn't know things that have happened after September 2021. So it doesn't know that Barbie crushed Oppenheimer at the box office. And for Sohan, two years is like a sixth of his life. So that really frustrates him. In the Future, we'll see LLMs being augmented by real time data to create a more dynamic experience. HubSpot AI assistance and chat spot are examples of this. They combine the power of an LLM with real time data from the HubSpot CRM. Now. Frustration 2 Today when we use AI, we mostly have to enter text as input into the prompt. In the future, we'll also be able to submit images, video, audio, code. If you've ever heard someone say, oh, AI is moving towards this multimodal future, that's what they're talking about. Frustration 3. AI doesn't really start doing things until you ask it to do something. It just kind of sits around passively waiting for you to type in a prompt. In the future, AI will kind of work in the background on your behalf. Now imagine hearing this message from AI one day. Now the voice you're about to hear is Dharmesh AI, which is totally not a thing. Listen. The number of new leads added last week went down 22%. Seems that one of your landing pages was changed recently and the conversion rate went Down. Also, I noticed that you don't have a chatbot on that page. You should add one. People are not looking to convert, they're looking to converse. Sounds almost human. A bit awkward at times. So basically it's just like me. So those are all exciting developments. But what has me the most excited is how all of these are going to combine to create what I think will be the next big wave in generative AI. Agent AI. And not the kind that wears sunglasses and a suit and a serious look. Agent AIs are AI powered software. They can work mostly autonomously to pursue specified goals. And they do this by acting as an expert in that given domain. And they are networked with large language models. With the Internet and and other agents, they are tailored not just for a specific use case, but for a specific user. Now, I know that seems a bit abstract, so let's look at a concrete example. This is a concrete example. Now, if you're wondering, for those keeping score at home, if this counts as a dad joke. Yes, it does. It's a multimodal dad joke. A real example. One of the most common needs many of us have is prospecting for new customers. Now imagine having a smart prospecting agent. It can search the web, it can crawl a company's website, all the webpages, and it can answer any questions you have about pricing, business model, whatever. So. And it would be tailored to you and your individual needs as a salesperson, let's say. So here's an example of how you might configure a prospecting agent in the future. Add five new software companies to HubSpot every day that are in Boston or B2B and sell a subscription based product. Notify me by email and include an analysis of their business model and pricing. Now, the agent goes off and works in the background and delivers on that particular goal. How cool would something like this be? Yes, yes, I think it'd be super cool. In the future, we'll have agent AIs for all sorts of things. It is super exciting. All right, that cue is for me to breathe. You're welcome to do so as well. Just so I don't pass out on stage. Slows me down a bit. All right, so for our final segment, I'd like to offer you some unsolicited relationship advice. Because the only time you should take relationship advice from me is if it's about AI. And to do this, we're going to pretend like you. The inbound audience are having a chat with Dharmesh AI, which is totally not A thing. So here's the prompt you might enter. You are co founder CTO of HubSpot. You have 30 years of experience in CRM and you are obsessed with AI. Answer my questions as best you can. As you wish. First question. How should I now be thinking about data in the age of AI? We already have a plethora of AI models and new ones coming out every week. Falcon, the open source one, came out today. 180 billion parameters. Super cool. By the way, I'd like to thank my wife for teaching me the word plethora when we first started dating like 30 years ago. It means a lot. So AI models are common. Data is the differentiator. Here's an example. Let's say you want to send a message to an existing customer about a new product you just launched. Your customer platform has the demographic data for the customer, every webpage they visited, every transcript from every sales call, every customer support interaction. You have deep contextual data. Now you take an LLM, pair it with that deep contextual data, and you can write the message that is the most useful and most effective for that individual customer. It's just like Yamini talked about. It's not just personalized, it's personal. So it's a bit like understanding the customer's love language. What makes them tick? What do they truly value? By the way, if you haven't read this book yet, I recommend it. Especially if you're in a relationship with something other than AI. Now, to truly drive customer love, you need to pair the LLM with a next generation CRM. So the first generation of CRMs were introduced over 30 years ago. They were good. Early cloud CRMs came out 20 years ago. Even better. And now we are entering the age of the Smart CRM. We launched HubSpot CRM at Inbound 2014. Time flies when you're having fun. Today, HubSpot marked the spark of the age of the Smart CRM. With a CRM that has AI included, we want to democratize this powerful new technology and help millions grow better. Okay, so a few of you are thinking, AI sounds awesome. Smart CRM sounds cool. But will AI take my job? You may not ask this question out loud, but a lot of you are thinking it. Here is. It's the most very common question, by the way. Here is the most common response you will get. AI won't take your job, but a human using AI will. If you're not using AI, I have a different take. AI will take your job. Now, before you take photos and slam me on Social media. Let me finish. AI will take your job and give you one that's much better. One that involves less cranking and more creativity. Less grind and more growth. One where you have superpowers with AI assistants and agents. Okay, now you can take a photo and tweet it. Or I guess it's exit. Now exit. Thanks, Elon. All right, so let's say you're bought in. Now it's like, okay, I get it. Let's do this. What should I do next? Depends on where you are in your AI journey. So I create a simple framework to help you kind of figure that out. Figure out what next step is. Phase one is awareness and investigation. Figure out what AI can do and try some things out. Phase two, Adoption Integration, Optimization. Use AI to get things done, weave it into your workflows, and then look for areas to get leverage. By the way, do you know what Old McDonald said when he saw ChatGPT? Stable Diffusion and Agents AI AIO. That's why I call this the AI AIO framework. Now, you may be wondering, would I go to the trouble of creating an entire framework just to be able to tell my old McDonald AI joke? Yes, I would. By the way, thanks for laughing at some of my dad jokes. It really helps to have a live audience. Like when I try these on Zoom, people don't laugh. Maybe they're not remotely funny. So remember, whatever it is you do, AI can help you do it better. Here's a list of some of the AI tools I use that make me better. And it's not just about being more productive or doing things faster. It's often about pushing open new doors. So let's say you're a developer like me, have always wanted to do some design stuff. Or you're a copywriter that wants to launch full campaigns. Or you're a salesperson and want to strategize like a CEO. AI can help push the frontier of what's possible for you. If you want to chat about what's possible, we're launching a new community today called AI Adventures. It's on connect.com. thank you. I'll be hanging out there all week, answering questions, asking some of my own. Hope to see you there so we can all learn about AI together. Last question. What advice would Ted Lasso give? What would he give me? Tell me about AI Now, I'm no Ted Lasso, but I am guilty of having binge watched all three seasons twice. Here's what Ted Lasso might say. I may not know the first thing about artificial intelligence. I'm not that into those computery things. But I'll tell you what I do know. Change is challenging. And new things can be downright nerve wracking sometimes. But if it helps you grow better, I say go for it. Because it's not about believing in AI. It's about believing in yourself. Barbecue sauce? Thank you.