AI - Powered Marketing
Artificial Intelligence is transforming the way we approach marketing on a seemingly weekly basis, so it’s crucial to stay ahead of the curve. Joe invited Clark Boyd, Lecturer at the Columbia Business School and author of the news letter hi.tech to discuss the applications of AI in marketing.
Joe: Hello and welcome to the Essential B2B podcast brought to you, as ever, by Lead Forensics. I am your host Joe Ducarreaux. This episode is the audio from a webinar we ran called AI-Powered Marketing, the future is here. For this I was joined by Clark Boyd, Lecturer at the Columbia Business School and author of the brilliant newsletter, Hi- Tech.
Clark is always worth listening to when it comes to the current state and future of marketing and as you’ll hear he has some brilliant insight into how AI can and will be used for marketing.
So without further ado here is Clark Boyd on AI-powered marketing.
Joe: What is AI-powered marketing and how is it changing the industry?
Clark: So, it seems like it should be a really simple question. I’ll try and answer it simply. It is something that enables a completely different relationship between marketers and customers through data. So, it is using artificial intelligence to make decisions on behalf of a brand, to gather data from customers and to cluster them together, figure out things about them;
What are the behaviours they show? What are the messages that they might be interested in? …..all happening automatically in the background. So that we as marketers, in theory at least, should have less of that hands-on work to do with the data. We can think a little bit more about the strategy and the creative side of things.
That’s AI marketing up until today. The reason that we’re here on this call, I believe, and the reason I’m having a lot of these calls at the minute, is because AI is starting to go even further than that. Up until now, what AI has really done is told us something about the world as it is already. It doesn’t necessarily interact with or change that.
It’s looking at all the data and surfacing something or reacting to what people are doing out there. What we’re now getting to with generative AI, as the name suggests, is AI that can create new content, that can essentially nurture customer relationships over time, in theory at least.
As we’ve seen with ChatGPT, that friendly user interface that has brought this to the public consciousness, it is able to create some pretty impressive things. So marketers are starting to think about what that could mean.
AI-powered marketing in the future will require a new definition that means AI can act as an agent on behalf of a brand, as well as someone that works in the background, telling us about what our customers are doing. The public face could even be an AI in future.
Joe: So I guess what you’re saying is up until now, it’s been an excellent tool at mimicking things that have already previously existed. But now we’re getting to the stage where it’s starting to create its own bits and pieces as well. There’s a little part of me that is slightly worried, just like, is this okay? Are we going to make Skynet? Who knows?
Clark: Yeah, because the way it is at the moment, the way that they would explain these large language models that it’s been trained on, is that what it’s creating is something that’s very similar to, but not identical to, things that have existed already. So it’s not completely thinking up, it’s not a Picasso kind of figure that’s just completely moving the whole conversation forward.
That’s where I think we maybe get a little bit too excited too quickly with it. What marketers need to do is figure out in real terms, what does this technology do? How does it function? But also, what does it not do? Where is it weak? Where does it create things that are plausible and seem convincing, but they’re actually false?
That’s where I think we need to be quite careful because it’s nowhere near the stage, I don’t think, of being this kind of sentient AI that people talk about going out into the world and everything.
But we’re willing to, I think, quite often accept a lot less than that from AI. We are impressed very quickly by whatever it can do. But in naked terms, what it’s doing is creating things that are similar to, just not identical to what we’ve seen before. Maybe that is creativity.
Joe: Well, I mean, we could get into a whole debate about whether that is valid for art forms. I think we spoke about that on a previous podcast. But you mentioned using AI to improve sort of customer engagement and relations, that sort of thing. So how can AI-powered marketing help to improve customer engagement, or even will it?
Clark: So I was thinking about this a bit before we started chatting. It’s one of the questions that keeps coming up. I think what’s so intriguing about this question is that we’ve had so much technological development in the last 5, 10, 15 years and yet we’re still coming back to this question every time. I think if we were to look at customer satisfaction 15 years ago at a base level and look at where it is today, would it be that much higher?
With all the technology that we have at our disposal, is that what we’re using it for? We should be, I think we want to but what we’ve lost sight of is what that customer satisfaction and what that engagement really is.
I saw during the Super Bowl, someone posting the old classic meme about marketers are from Mars, consumers are from New Jersey. Now, we’re thinking about all these crazy ideas we could do, but they’re thinking, will there be free parking at the event? What’s the weather going to be like? Ah, I’m not going to that.
But of course, we’re sitting in a room drawing up personas and saying this really hits a touch point for this consumer. It’s not really the way that they’re thinking. So AI could lead to better customer engagement and satisfaction if we get to the heart of what those points really mean and understand them.
Because we want to create an AI generated video that’s personalised for every customer when they land on our website. That might increase the loading times by seven seconds, so they don’t even stay to watch it, but it’s something we spent three months on.
I know this because I’ve done it. I’m not, when I say we, I’m not talking about any particular person out there that I might be insulting. It’s me. I’ve done all of these things and I understand why, it makes our job more fun.
But actually, I can be a part of helping us understand that. That’s where I would draw this question back to. It’s not just a human question for us to think, what does it mean about customer satisfaction? Actually, what some companies are doing that I think is really clever use of this technology is they’re using it as a new layer.
So the customer comes to the website and they interact with it and they bring in those data points to understand what it is that this customer is really looking for? Maybe they interact with a chatbot. Okay, will this require human intervention? Is this going to be a serious issue? Will this person be more annoyed if they have to deal with a chatbot, no matter how smart it is, because they want to speak to a person. So they’re bringing it in as a kind of filter and then it triages all of these different kinds of complaints through the website.
So that’s something that can be really effective. It’s not going to win a big creative award or anything like that, but AI can help us understand customer engagement and then deliver better on it in the future. I think that way of thinking about the question will yield better results.
Joe: And that, Clark, in Microcosm is entirely why I wanted to speak to you about this, because there you are already thinking of the bigger picture to do with the area of marketing. Can we just get your thoughts on Google’s announcement about Bard.
Clark: I’ve been strangely quiet on this one. Normally you can’t stop me mouthing off about every announcement under the sun. I’d normally have a newsletter edition out about all of this sort of stuff and this one came out and it just felt like literally everybody was having their say. I like to put these things in the wider context, as you say. If you look at what’s been going on with Google up to this point, it’s certainly not surprising that they’re going in this direction. The surprise is the speed of the change that they have made.
They have been rushed into making a change in a very, their employees are calling it, an un-Googley announcement. But it didn’t have that impact or that slickness even that people are used to, where it’s just rehearsed and rehearsed and rehearsed. They wanted to get this out very quickly.
Now, if you look at what’s going on with Bing this week, there’s been so much going on there. There’s a whole subreddit about the mistakes that its AI is making, but that isn’t affecting Microsoft’s stock price because Bing is a tiny search engine. If Google does one thing wrong, it bites $100 billion off its market cap.
So that is, that’s the lens through which people at Google are thinking about all of this. It’s not how do they make sure they’re not caught on the hop. What they have announced though, I don’t think is as revolutionary as people are making it out to be. There’s a lot of talk about, search will never be the same again.
This is completely radically changing everything, and I don’t see it. I think that’s the main reason I haven’t had much to say on it. Because I don’t see what is so radically different about it.
So to bring it back to what they’re announcing with Bard, a terrible name it’s almost as bad a name as ChatGPT, but what Bard would do in Google search is that it would act essentially as your personal assistant. So you could ask it to plan an itinerary. I’m going to whatever place, I’m going to a Greek island in July. What’s going to be on? This is what I’m interested in. Go and find me some activities.
So instead of me Googling all those terms, going through annoying pop-ups and clicking through, trying to find the little snippet that I’m looking for, it goes and does that for me. But that’s what Google wanted Voice Search to do. If we had all used Voice Search more, that’s really what it wanted it to do. It’s been calling search conversational search for quite a long time now, because you can type in, who is the President of the United States? Joe Biden. And then you can just type in who is his wife? And it’ll go on, yeah, Jill Biden.
So it’s already doing a lot of these things in the background. It’s not bringing it into the front. That’s the bit that Google has been rushed into that it didn’t want to do, because it has this perfect profit-driving machine of just tell us exactly what you want and then we’ll tell you our best guess of 10 things you’re looking for. But before you get there, here’s seven paid ads that we’re going to charge a fortune for because they know exactly what you want.
That is much harder to manage if it’s very dynamic, and what if it’s just one query and it’s just you and the chatbot? So I really see Google having this as almost like a widget. It’s an option. There’s a chat option within search that you can use and they’ll figure out a way to monetize that as well.
From a marketing point of view, people always think, well, what does this mean? Does it mean Google’s going to steal our content? Well, Google’s been stealing your content for years. That’s something people have put up with. People were up in arms about it, and rightly so, and they went along with it because they had no real option, it was their means of getting to customers.
Google steals your content to answer questions all the time and it doesn’t credit you. With this, it is supposed to cite its sources. Maybe that’s the best we’re going to get. We can hope that it’s just certain types of queries. But the more interesting bit, from my point of view anyway, is that these chatbots are going to be offered as part of e-commerce experiences, so on websites.
Not just Google as an intermediary. It doesn’t want to be that anymore. It wants to be right in the website. Bing will offer this through Microsoft Azure services.
So every company would be able to train and build their own version of Bard or ChatGPT. Just depending on who you’ve chosen as your cloud provider, that’s the chatbot experience you’re going to get. So then you get a search experience on websites where the chatbot can go and find images for you and suggest outfits for you and do things like that. That’s the understanding of Google that we need to have when we think about this conversational chatbot experience.
So on the whole, they’ve been rushed into this. They didn’t want to get caught by Bing. The companies that have dominant market positions just panic all the time about somebody coming in and taking their market share. So they’ve been panicked into making this change. It will be an option if you’ve used the Bing one so far, I know there’s still a wait list for that one. It’s just an option you can use. It’s not always that useful.
Where I do see it happening is onto websites, into things like social media, connecting with your database as a company and your inventory. That’s where Google wants to be because at the minute it’s losing out to Amazon on that front and that’s the bigger competitor, as well.
So Bard is not just about search. It’s really about something integrated into fighting off TikTok and Amazon and Bing and all of the competitive threats that Google has.
Joe: And yeah, and as you say, panicking to the tune of, in the article that I read, to the tune of $120 billion.
Clark: Just for one little mistake! I mean, it really shows you how much they have to lose there, doesn’t it? Make a couple of little errors and that’s enough for the market to think Google’s lost it.
Joe: Just circling back to the ChatGPT functions and implementing AI in marketing strategies. We’ve touched on a few of them, but can you list some more? What are some of the challenges that businesses face when implementing AI into their marketing strategies?
Clark: So, there are quite few and it’s interesting because the big one with AI, I’ve been talking to companies about this for the last number of years and the big problems are often technical implementations and the complexity that comes with bringing this kind of technology into your everyday. That’s no longer really the case.
If companies were slow to do all of that, they’ll be slow to implement this too. It’s not going to make them faster. But when they make the decision to implement this kind of technology, it’s much more straightforward than it ever has been. It’s kind of plug and play in a lot of areas.
So the question then becomes, how do you ensure that this AI gets the kind of results that you want it to, that it acts responsibly on your behalf. Because companies like Microsoft and Google are making the mistake of letting this loose, because they want to be first to market. But the Microsoft one is going a bit crazy already.
People are finding ways to hack it, to get it to act as an evil agent and do all of these crazy things, and it’s not taken them long. So as a company, you don’t really want that technology doing whatever it wants on your behalf, do you? It’s a question of accuracy. It’s a question of quality as well.
There’s a lot of testing to do before you would put anything like this live. There are lots of ways that you can modify it and get it to work for your purposes. There’s going to be a big market booming in that over the next couple of years at least.
But the thing I would ask companies there is, the cost of creating content has been driven down to zero, but the cost to your brand reputation could be absolutely huge. Because we’re not always thinking about the customer at the end of this, it’s the problem we always face here.
We are really excited by the fact that we can create lots and lots of content, we can create images for every individual customer. Does that customer want to see those images though? Well, we haven’t really thought of that. The content that it produces is pretty good. Well, how often is a brand happy to put pretty good content out there? It shouldn’t be happy to do so anyway.
We spend a lot of time thinking about these things and we spend weeks and months even just writing a blog post making sure it’s just right. So are we going to switch from that to just hitting a button and it creates 500 words that we promote to every single customer? If it’s personalising this to every customer, how will you vet that?
So there could be thousands of people seeing different articles on your website. Where’s the brand identity and brand control over any of that? How do you vet any of it?
So there’s still a role for people, call them humans in the loop with AI systems, where you’re giving feedback to it, you’re understanding what happens, you’re actually assessing and analysing the quality of its output, not just setting it live.
But it’s interesting because for businesses, this is a new way of thinking. Up until now it has been, how can we possibly use AI now?. It would take us a year or two to implement an AI system and then would it even really be AI? And now, companies like OpenAI are saying, well, here you go, do whatever you want with this.
You can use this API, it’s going to be very cheap and do whatever you want. Companies can ask very different questions based on that nod, that I’m not sure they’re ready for, but it is things about quality, accuracy, even copyright ownership. I mean, what if one of those blog posts used a piece of artwork, but amended it and called it an original piece by your company.
The artist sees it and then takes you to court. You didn’t even know what happened, but the AI is acting on your behalf.
I’m certainly no legal expert, I’m not going down that rabbit hole, but you can see how that can become a real challenge. Now, brands are actually having to switch all of this on its head, not AI is here and we can use it, but actually, hold on, should we? It’s a completely different position.
Joe: Yeah, I really like that in all your fantastic answers, what that one essentially boiled down to was Jeff Goldblum in Jurassic Park. You didn’t stop for a second to think about whether or not you should.
Clark: Just because you could, you just went ahead.
Joe: So, Clark, can you give us some tips on how you would get started using AI, particularly for a B2B business, at all?
Clark: So there are lots of good ways into this. It’s one of those strange situations, it feels like you could do anything, so you end up doing nothing. Because a lot of the things that you read will say, it’s just down to your imagination or literally anything is possible now. There are a lot of people thinking, well, I haven’t really played with this kind of tool before. I don’t really know where I’m going with it. I’ve actually found it, I think when I first got access to ChatGPT, I thought there’s nothing to ask this thing.
What do I want to know? Then I started just asking the questions, like who’s going to play at left wing for Leeds United Saturday? and it didn’t know that. So I was like, oh, okay. I should be thinking about these things more. I was like, okay, so it’s not like a magic eight ball. It doesn’t know anything about the future. How can it help me? You probe around a little bit.
One of the things that I find really helpful with it is, as marketers at any business, whether it’s B2B or whether it’s small, whether it’s a huge business, just as individual marketers, trying to think about how this technology works, there are two big ways to do it.
The first, and certainly the easiest way, is to go into say, ChatGPT, go into Dolly2 or Stable Diffusion, the image ones and think about something that you would like to do, something you would love to achieve. It could just be, say, you write a newsletter…..actually, yeah, maybe I should be using it! You’ve got a newsletter and you’re trying to think of five angles for a story.
So, you just put in, start off really basic, start with the shortest version you can. Say, brainstorm five ideas on this topic for my newsletter. Look at the output and then you’ll see, okay, so it doesn’t understand my audience. Doesn’t know anything about that. It doesn’t know anything about my newsletter. Those topics are way too broad.
So you might go in and refine it and say, I write a technology newsletter for marketing executives. Come up with five ideas on this topic, but go beyond the obvious cliches and then use the examples that it gave you. It’ll go a little bit weirder with them after that.
You can even say, well, are those popular topics at the minute with my audience? It might go and fetch you some information about those. You can go down that path and see, what are the little modifiers in the prompt that will help you understand how it’s working in the background. Because then you can get a lot more done with this and you can come up with template prompts that you use and you can share with your team. Right, if you’re looking to brainstorm, say this, this is how it understands who we are and what we do, and it’ll get you the results you need.
The second thing though, and I think this is much more powerful, I think this is what has really, really helped me build, using these tools actually, is research the APIs. So they have documentation on the APIs, on OpenAI and it is very, very clear. Gives you lots of use cases.
It says you can use the technology that powers things like ChatGPT and like Dolly. It’s open for you to use as well. What those tools are, it’s just a user interface. ChatGPT is a slightly different API that isn’t open to the public yet, but it will be soon. What they’re doing is, they’re essentially wrapping it up. We call it an API wrapper.
It just wraps the API call into something that a user with no programming experience can just type in, oh, how do I brainstorm five ideas? If you’re only using it that way, the ideas that you come up with will probably be let’s make a chatbot or let’s make a content creation machine or whatever else. If you think about it the other way around, which is, I’ve got a challenge, I need to solve it.
Could OpenAI help me get there? Then you can look at what the APIs can help you do and figure out new ways to get to that point. So, what I would advise people to do as a way into this, beyond reading just that documentation, it’s short, it’s easy to use, it’s step by step and anyone can just get started.
It will generate an API key for you and then you can go away, input that and if you get stuck, do what I do, just ask ChatGPT. I’m trying to do this using your API. It’s not working. How do I fix it? It will literally tell you exactly what to do. If it doesn’t work, come back and just keep trying.
I have it open all day and if I hit a snag with coding, I will just copy and paste the error into ChatGPT. It will write code that will get me out of that error. That’s how helpful it can be. But before you dive into all of that, look at your day-to-day work. Look at what you’re trying to do for the next quarter, for example and just pick one thing.
So it could be the emails that you’re writing. Could you get those drafted using ChatGPT and pull that into your Gmail so that it would be there and you could just edit it and send it? Or do you need to research 10 new prospects? Okay, how could you use Google Sheets with the OpenAI API to go and fetch information about these companies?
Because you could just type in the name of the company, hit enter and it would go and find everything you need and populate the spreadsheet for you. So all of these things are really possible at the moment, but the best way to get into it is think of something you really want to do and then you will be curious enough to get to the end point.
You can look at my LinkedIn and my Twitter. I’ve shared a couple of things I’ve built just in the last two days with interactive marketing personas and image generation and things like that. Just for training purposes so I can show people how to use these APIs. But it was amazing for me working through that process.
It made me realise this is the best way in. Pick a problem, use ChatGPT to help you out, read the documentation and then you’ll be able to build something. You don’t need to code, you don’t need to do anything like that at all. You just need a bit of curiosity and an idea.
Joe: That’s fantastic. Clark, thank you so much for joining us for this webinar all about artificial intelligence. I must reiterate, go and have a look at Clark’s newsletter. It is fantastic and you write creatively and humorously about all sorts of topics. Clark, thank you so much for joining us.
Clark: Thanks everyone
Joe: Well, there you go Clark Boyd they’re joining me to chat all about AI. Our key takeaways then:
– AI marketers should understand what the technology does and doesn’t do and be aware of where it can be weak and create false information.
– AI could lead to better results for marketeers as long as they understand the true needs of their customers and use AI to address those needs.
– Chatbots will be offered as part of e-commerce experiences and companies can train and build their own version of Bard or ChatGPT.
– You can test AI by picking one task and seeing how ChatGPT can be used to make it easier. Remember Clark’s example using the coding.
Thank you very much for listening to the Essential B2B podcast and thank you to Clark for joining me. Please remember to subscribe and review Essential B2B podcasts wherever you get your podcasts and give a 5 star rating where possible. We’ll be back next week with another episode of the Essential B2B podcast.