How Can Marketers Use Artificial Intelligence

You buy things that are AI powered for the same reason you would buy traditional marketing technology. It needs to be solving a business challenge you have, or it needs to be helping you get to a goal more efficiently. It just so happens that there's now smarter technology on the market.

So, Paul, can you tell us about how marketers have used artificial intelligence today?

Yeah, we've seen a number of ways. Like, we've actually had one company that was about a $10 billion manufacturing company, literally just came to us and said, we what would it cost to do AI like to be AI first as a marketing organization? And I said, what in the world does that mean? And they didn't know. And so you realize that because it's sort of the buzzword now, and people just hear AI and they assume like, well, we need that, whatever that is. And what we always advise is like, at the end of the day, it's just smarter marketing technology. Like, you buy things that are AI power for the same reason you would buy traditional marketing technology. It needs to be solving a business challenge you have, or it needs to be helping you get to a goal more efficiently. It just so happens that there's now smarter technology on the market. And so as a marketer, you need to just know how to determine, is it really using machine learning? And is that machine learning making the product better? And is the product roadmap for that company building more intelligent software that will do one of two things for you, if you're lucky, both make you more efficient as a marketer and as an organization, and help you drive revenue or achieve your goals more efficiently. So that's really what it does. And so any way you would use AI, you're going to go through the same bot process you would go through if you were buying marketing automation software, email marketing software, conversational AI. You're looking at a business case for it and you're trying to prove out what the value will be to invest in it. I mean, how do you guys. When you're roadmapping, which features in HubSpot that you're going to integrate machine learning into? Like, how are you guys thinking about it?

Yeah, there's a few different types of use cases that we look for, recommender systems. So that could be blog content recommendations. We have a module for that on the CMS to recommend relevant blog posts based on the current blog posts that you're viewing. Optimization problems. So if. Which can often be like personalization. So with our lead flows tool, we have a feature called continuous optimization, which is going to adapt how that lead flow is presented based on what's on the page and based on other offers that that person has converted on. And also categorization problems. So things that can help you identify what is the position potential customer problem that's in the support case. So in our tickets object for categorizing types of topics across those tickets. So recommendations, optimizations and categorizations are three broad use cases that you can apply. Supervised learning, specifically that businesses can start to look at today.

Supervised learning being giving it data to train it on and then teaching it yes or no. That was like a good action, good response, good recommendation.

Exactly. So you mentioned earlier historical examples. So you need to know in this situation what was the human decision that was taken. And you can use those historical examples to apply that without humans to a wider range. So you can take a little bit of human intuition and scale it very quickly to identify high priority deals or high priority or good fit contacts. Urgent support cases.

There's a lot of email deliverability. I know it was an early application for you guys. How do you get this through the spam filters? Is this good email? Is it a bad email? What are the factors that go into it?

Exactly. So we have a team at HubSpot focused strictly on deliverability and they've been manually Reviewing sends that HubSpot customers have put out that have a particularly high bounce rate. And based on those examples, we can kind of map those characteristics to other sends. And so as we're sending even more and more email, it's helping our deliverability overall increase.

One of my favorite examples that's really tangible to most people when they're thinking about AI is Gmail. So the idea of training, so if you think back over the last couple years when you're using Gmail and it would recommend the three to five word responses, which was like the smart reply. Every time you as a user was saying yes, you were saying this is a good response and you were just choosing that three word reply. We were training their algorithms that that was a good recommendation. And now we see within Gmail, you know, fast forward a couple years, Smart Compose, where it's now recommending entire sentences to us. So it's looking at the sender, probably everything you've ever communicated with that person, the subject line, any context within the emails itself. And it's trying to make predictions about what you want to write next. So I mean, Google's always obviously kind of ahead of where the rest of the market is getting to because they have a massive team. But even like, Microsoft has, like, 3,000 machine learning engineers now working on solving things. So, yeah, I think at the end of the day, it's designed to make our lives as marketers more efficient so we can maybe focus on more human, unique human. Human abilities, I guess.