Author: Hannah Wehrly

A Napier Webinar: Uncovering the Truth about Artificial Intelligence in Marketing

Artificial Intelligence is the latest buzzword, but to what extent does AI truly make a difference to your marketing?

Napier recently held a webinar 'Uncovering the truth about Artificial Intelligence in Marketing', which explores the true impact AI has on marketing activities. We address:

• What marketing tools vendors mean by AI
• The truth about AI in B2B marketing
• Examples of marketing tools that use AI
• How your B2B marketing campaigns can really benefit from AI
• The future: how AI will change the marketing landscape

Register to view our webinar on demand by clicking here, and why not get in touch to let us know if our insights helped you.

Napier Webinar: ‘Uncovering the Truth about Artificial Intelligence in Marketing’ Transcript

Speakers: Mike Maynard

Hi, and welcome to our latest webinar from Napier, where we're going to talk about the truth about AI in b2b marketing. And so AI is obviously a very hot topic at the moment. But what we want to investigate is in reality, you know, how much impact is AI having a marketing today? And how much impact will it have in the future. So we're trying to skip past some of the marketing claims that we see from some products and look at, you know, actually, what is the real life impact. So hopefully, you'll come away from this webinar, understanding a little bit more about AI, how it can help you today and how it might change b2b marketing in the future.

If we look at the agenda today, we're really going to try and get to the you know, the bottom of what the truth is in terms of AI. So we'll start off looking at the marketing technology landscape, which is clearly a very complex and confusing landscape with lots of different vendors. We're talking about, you know, what AI actually is, and what we mean by AI, there are actually different sorts of AI. And depending on which particular type of AI you are considering, you can take away very different conclusions about how much AI is being used within b2b marketing. We'll talk about the applications of AI. And also the tools that integrate AI today, into their marketing systems will then go on and say, well, could you create your own artificial intelligence, that actually helps you directly with creating b2b campaigns. And within that section, we'll actually talk about somebody who we know who actually did that, very successfully, we'll look to the future, and to find out how AI might change the marketing landscape in the future. And finally, as always, we'll give you some top tips from Napier. This time, it'll talk about how to benefit from AI, both now and in the future.

So this is a massively complicated chart, produced by Chief Mar tech, a website that looks at the marketing technology landscape, they've been looking at the landscape for several years now. And they've gone from, you know, literally a few 100 marketing technology companies to this incredibly complex landscape with 5000 companies and 8000 solutions. So this is a massive increase. So clearly, there's a lot of companies investing a lot of time and effort to building marketing technology tools. And of these companies, very many of them actually are claiming to use AI within their different algorithms. So what we're going to do is take a look at, in reality, how many of these tools actually use AI.

But before we go ahead and talk about that, we need to be clear about our definition of AI. And it can mean very different things. And the two big areas that I'd really put AI into are Firstly, algorithms. So this could be formulae or this could be, you know, logical decision trees. And basically, it's a fixed set of rules to actually optimise a campaign or to do something else as part of your b2b marketing activities. And the key thing about these algorithms is they're programmed by a human. And they'll typically be fairly fixed, or they'll have very limited ability to develop or learn. The other side of AI is machine learning or neural networks, and machine learning is completely different. Now, if we look at machine learning, this is really computers, you know, understanding the world, and then trying to apply rules based upon the understanding they built. So the key thing about machine learning is you need to train the computer with various data sets. But once you've done that, the computer can do things and have insights that a human may not have. So potentially a computer can do more than a human. And this is certainly the most exciting part of AI, although today probably not the majority of where people think AI is being applied.

And then finally, AI can be the specialised or broad you talk about narrow or general AI. So narrow AI tend to be focused around trying to achieve a particular outcome, or perform a particular activity, whereas general AI is basically your howl in 2001. The expert computer you speak to as though it's almost another human. And of course, when we look at Mark marketing and b2b marketing, we are going to focus on narrow AI. So we're gonna focus on AI that's designed to do certain things within the marketing mix.

And if we look at Wikipedia, I mean, Wikipedia says artificial intelligence is an intelligence demonstrated by machines, but it's unlike natural intelligence displayed by humans and animals. And to me, this is interesting. So, um, it's very true that AI isn't quite like humans, animals, intelligence. If we look at you know, the algorithmic type AI formula that have been programmed in, that's not really what we think of intelligence, we think of that as applying rules. and machine learning is very different. It's really pattern matching on a, on a massively complex scale. So machine learning, it can give the impression of being intelligent like a person. But it's still is very different from the way we work. Although the structure of neural networks are somewhat modelled on the structure of the brain,but whatever we do, we're not seeing something that's going to directly replace humans, at least not in the foreseeable future. And I think that's an important point is whichever type of AI we're doing, whether it's broad or narrow, you know, today, we're a long way away from replacing people, we can certainly help people make them more efficient, make them more effective. But I think replacing people is is certainly a long way off.

So let's look at the definition we're going to, to have Oxford languages has a definition, the theory and development of computer systems, able to perform tasks normally requiring human intelligence, such as visual perception, speech recognition, decision making, and translation between languages, we're going to actually take something somewhat like that and apply it to marketing. So our definition is a computer algorithm that can learn from data to produce insights and recommendations specific to the brand or campaign. So what we're saying with AI is that really to add value, the algorithms got to learn, it can't be a pre programmed set of rows.

And it's got to produce things that are specific. So if it's learning from the activities for a particular brand or a particular campaign, it should then be able to produce recommendations that are uniquely beneficial to that brand, or that campaign. And so we're gonna move forward, we will talk a little bit about algorithms.

But in particular, what we're going to look at is where we are today, in terms of the state of AI, as relates to marketing, from the point of view of learning from data, so this is really the machine learning the neural networks. And so our goal today is really to look into whether there's been a huge impact from machine learning on b2b marketing.

So the first thing to say is clearly there is AI in marketing. In fact, there's the marketing artificial intelligence Institute. And that is an institute dedicated to promoting the use of mark of artificial intelligence within marketing. So clearly, there's there's a lot of AI going on lots of marketing technology tools, promoting their AI functionality. And actually, the CMO survey, which has been running for over 10 years, in 2019, show that almost 60% of the respondents say they're using AI today, although the main uses of AI are in personalization of content, and predictive analytics. And in both of those areas, they tend today to be algorithmic type applications. So you might see the computer delivering different content, depending upon, for example, which persona they think somebody belongs to, or which company they belong to, if you're doing reverse IP lookup on the address, or maybe even what pages they visited, but it's a set of rules. It's not necessarily this learning machine that we talked about. So it tends to be simple algorithms today. And I think that's really important. Because although people are using machine learning, really to get on board today, the bar is not very high, to be able to use some of the the tools that are available is relatively straightforward. You know, predictive analytics can be as simple as lead scoring algorithms. So at the very simple level, I think everybody can start using AI in the most basic sense. And actually the more advanced level of machine learning level, as we'll find out, most people in general are not really using machine learning for things that are specific to their campaigns. There are areas where machine learning is being used, but it tends not to be campaign specific. So if you're not using AI today, the message is, you know, get on board, don't think it's too difficult. You can use some very simple approaches to improve your marketing campaigns to reduce the amount of time you spend working on managing the campaigns, and hopefully also to get better results. And we'll talk a little bit about this as we go through.

So let's look at some of the simple examples of AI in marketing. The first one is content creation. And one of the most common things you'll see is email subject line generators, and typically you either give it a benefit or a product or a service. click the Generate subject line and magically you get these amazing AI generated subject lines. And you can see on the right of the slide here, that we've actually tried it. And we've tried it in terms of, you know, learn about marketing, learn about AI marketing. And you can see the first subject line that is suggested actually isn't great grammar, how to take the headache out of learn about AI and marketing.

The others read pretty well. But all of them use a very simple approach where a stock phrase is put in front of whichever benefit you have. So, you know, if we look at the lazy persons way to learn about AI marketing, it could be the lazy persons way to write an email to create great email subject lines, you could put anything at the end there. So although this has been put out as AI, again, it is a very simple, very formulaic approach. And frankly, you know, if you're stuck, it's not a bad place to go for an idea for a subject line. But I can't see this sort of AI replacing copywriters anytime in the future. You know, to me, I think it's going to be sometime before we get things like subject great subject lines written by AI eyes. And one of the things you do see with AI is actually some AI tools that can grade subject lines or give you an indication of which one's likely to perform better. And that might be more useful. But still, it's the actual copywriting still needs a human to write great and innovative subject lights.

Content personalization, one of the key things that was mentioned as a use of AI today, and many tools offer content, personal personalization, that might be your marketing automation tool, it might be a content management system. Or you might have a specialist tool or a plugin that enables content personalization. And most of them are based on fairly standard fairly simple fixed set of rules. So it could be on the persona, the company, the roles, something like that. And actually, you know, from my point of view, the most important thing is the insights that are used to personalise the content, they're generated by humans, they're not generated by the machine that took the tool that you've got, the machine will automatically provide the right content. But it's got to be a human that works out why one persona needs a particular message. And another persona is a different message or a message phrased in a different way. So again, although there's some automation here, it's still very algorithmic, you know, it's very much a formula, if this person is a CEO, then talk about financial benefits. If this person is an engineer, then talk about technical benefits. So it's very different from something that's that's truly autonomously intelligent. And again, you know, it's a great way to get into AI. It's a very basic first step.

But we're still some way away from this, this idea of computers actually running your marketing campaign. And certainly, anyone who's used these content personalization tools, like Martin automation, will now how incredibly effective even the most simple personalization can be. So it's a valuable tool. Even though you know, from our earlier definition of AI being machines learning, typically, most people are not using machine learning tools.

So these are the simple thing is, why don't computers learn? And the answer is, it's actually quite hard to train a computer. If you talk to your friendly data scientist, they'll talk about, you know, building a neural network, and then building a training set. And the training set is the data that's used to basically train that, that algorithm.

And the first question everyone's gonna ask is, you know, how much data Do I need to get a computer to learn? And the answer is, Well, it depends, like many other things, it depends a lot, it depends on how complex the problem is, how much data is coming, and how many answers are coming out. And it obviously depends on how accurate you want the algorithm to be. Actually, if you look at these machine learning algorithms, they will have a certain tolerance for error.

And depending on how much error you're prepared to accept, that will affect you know, perhaps how complex your your network is, or how big your training set is, or both. So there's no real answer in terms of how much data you need. But typically, it's 1000s of data points. And so if you look at, you know, some of your marketing activities, you may not have 1000s of data points, you may not have 1000s of email subject lines you can compare to find out which one's going to be most effective for your audiences, for example, I'm making it even more complicated is that the environment changes. You know, a great example would be that, if I a year ago put the new normal into an email subject line, and that would at least be fairly unique and probably be fairly confusing to people. I mean, today

The new normal seems to be on every other email subject line I receive. And it means something it has context. Now, it didn't have the same context a year ago. So with a continually changing environment, you've got to be careful about going too far back to get the data. So things that change in the environment might be, you know, the way people speak, there might be, for example, means that come in. But also, you might produce new products, new services, or have new markets, all of which will change what works for your audience. So you've got to be very careful about going too far back. And this makes it very hard to build great training sets for b2b marketers, because it's very hard to get enough data that is of sufficient quality.

I mean, so this is the bad news. You know, computers are pretty slow learners. But the good news is, is they're much less likely to jump to the incorrect assumptions that we often make as humans. And the number of times I've seen people look at, you know, email open rates, or click through rates on Pay Per Click ads. And the human immediately jumps and says the one with the highest percentage must be performing better. Whereas the mathematicians will say, No, no, you got to look and see whether it's likely to be randomness, or it's actually likely to be because the ad is performing better. And typically, I find that most people who are not familiar with the idea of statistical significance, they'll actually jump to conclusions well before one ad, or one subject line has proved to be more effective. And quite often, you can keep running the ads for a longer time and over a much longer period, you'll find that their assumption was actually completely incorrect. And that the other subject line in the long term performs well, but the noise from randomness meant that, you know, early on in the campaign, the poorer subject line looked like it was doing better just because, you know, there's a random chance of whether or not someone will open an email. So they don't jump to conclusions. And certainly the data, scientists will make sure that they create the models and the training sets that stop computers jumping to conclusions too quickly. But they are very slow learners. And this makes it very hard to take advantage of machine learning.

And if you look at it, actually, there are a limited number of applications of AI, that really, really make a difference in marketing. And I kind of put them down to three, this this Smart Insights diagram also has kind of, you know, different areas. They look at propensity modelling how likely someone is to buy. that's fundamentally the idea of scoring leads or predictive analysis, it looks at dynamic pricing, something that actually isn't so relevant for b2b. Typically, b2b doesn't use dynamic pricing pricing is fairly consistent, and often negotiated. And finally, predictive customer service. Now, for a lot of b2b companies, they've been using predictive maintenance or predictive customer service for some time, where they're monitoring the performance of a system. And where they can see the system is potentially going to either have reliability problems, or perhaps you know, if you're thinking about a system of a certain capacity, maybe run out of capacity, then the company can preemptively phone the customer, or contact the customer and say, we can see there's going to be a problem. And here's what you need to do to avoid it actually happening. So kind of predictors customer service already exists. In reality, if we look at the applications of AI for b2b, it really forms into three key areas around marketing itself. So that's look alikes. So that's saying, you know, if I know this particular profile of customer, tends to be a good customer for me, then find me similar, similar people who might well be good customers, and people who've used Google ads, for example, would have almost certainly use look alike audiences. For example, you might want to target people who Google believes is similar to the people who visit your website. Very simple use of look alikes great use of AI and something that can be very, very effective.

You can use AI in terms of predicting intent, and that can be as simple as lead scoring. Or it can be a much more complex algorithm that learns particularly on e commerce sites, you know, who's likely to buy and if you buy one product, what other products you're likely to buy

And finally, performance prediction. So this is being able to assess whether a particular campaign is like to perform well or not. And these three areas of AI are all the things that you can actually use now that will use machine learning and help you with your campaigns. But as you can see, they're fairly small elements of the overall campaign. So today, it's very clear the AI is not taking our jobs, it's there to help us. And there's some way to go before AI takes over a whole campaign. Interestingly, marketers also use some other AI applications that are on the face of it completely unrelated to marketing. So natural language understanding is widely used in things like chatbots. And image recognition is widely used in digital asset management systems. And we'll look at some of those in a minute. In one of the next slides.

Interestingly, though, if we think about this, you know, I find that you know, things like look alike audiences or predicting intent, they're probably the least exciting forms of AI. But actually, that's where everyone's getting the biggest benefits. So although the message that AI can do everything, and it can write content, and it can optimise content, it's all true. Actually, the reality is, is very few people get benefit from those. And most people get benefit from rather simpler applications. And again, this is great news. If you've not widely used AI, it's very easy to get on board and get up to speed very quickly, the bar today is is relatively low to, you know, start using AI and benefiting from it.

So firstly, the good news, you're probably already using AI, I've mentioned this before, but pay per click advertising often uses AI. So if you're using Google ads, you might be using smart bidding, which will basically determine how much you pay for an ad, using an AI, you might be using responsive ads, which will actually determine what content is in the ad. So the headline and the description, based upon performance, that's again, using AI, you might be using look alike audiences, or you might be running the Google smart campaign. So all of these can use AI to help your campaigns work better on Google. And it's not just good. I mean, Facebook, as well as another great example that's got some powerful AI and its advertising tools. So very simple way of, you know, using AI is to get on board with Google ads, and to start using some of the ai ai functionality there. And it's interesting, Google's actually motivated to make the AI work really well. Because clearly, the better results you get, the more you're likely to spend with Google. And so it is in Google's interest to make sure these API's really deliver the greatest results, because they believe longer term that will increase your total spend. Now, of course, there is a risk of the commons being affected by this. So with a limited number of searches that are available to bid on, of course, you know, if everybody starts increasing their spend, and the cost per click is going to go up and Google is going to be even happier. They're not just getting more revenue, they get more revenue per click. But either way, you know, once Google releases this, you should be making use of it because it's optimised to give you the best results, not necessarily in the short term to give Google the best results, because Google reckons that your short term gain is their long term benefits.

Chatbots AI, another area where people might be using AI now, we all remember, I think we all call him Clippy, or I believe that Microsoft officially called the little paper clip, clip, clip it and could be used to pop up and say, all sorts of unhelpful things whilst you're trying to create a document. And it was a first attempt at a kind of chat bot, you know, you had effectively buttons to click rather than being able to answer it in natural language. But it was kind of an attempt to provide some level of AI. Now today, what we're seeing is that with web chat widely used, chat bots are becoming more and more common. A lot of enterprises are looking at chat bots. And by the end of this year, Oracle thinks 80% of enterprises will use chat bots. And the reason for this is a lot of inquiries on a website, are actually inquiries that can be dealt with fairly simple, simply, and it can be automated. But obviously what we need is we need to understand natural language. So you know, even asking for a bill for a mobile phone, for example. You know, people could ask, Where is my bill? How much is my bill? What's my bill? And so you need some intelligence to process the way people ask questions and the different ways they can ask questions. And so natural language processing is key. And that's an area where AI really excels. And of course, one of the reasons I really excels Is there a vast training sets of natural language that people can use to train API's.

So we see chat bots becoming better and better as people use them more. And actually, if we look at another big trend is that voice technology particularly Alexa is becoming more and more common and by the end of 2021 it's forecast that 40% of companies will adopt voice technology. So they'll be creating things like Alexa voice skills by the end of next year. And the great thing about this is whether you're writing a chatbot or using Alexa, you don't have to do anything to understand the question. The tool you're using the chat bot tool, or the elixir API does all the difficult work of understanding what the user is asking. And all you have to do is then create a series of simple rules based upon the kinds of questions you get. So it's very, very easy to generate these chat bots and generates voice skills. And it's something I think is going to grow rapidly over the next couple of years.

I mentioned image recognition. This is this is interesting. You know that there are huge data sets available for images. So it's relatively easy to train. AI's to understand images, not only in terms of what's in the image, you know, whether it's a for example, a ski boot, or a stiletto heel, but also to understand things like colour, and even understand facial expressions. So whether someone is happy or angry or frustrated, and the great thing is people building these generic, AI's can then have them applied to your particular image library.

So you can categorise products. And you can also look at sentiment as well in images. And one of the biggest is Google Cloud Vision API, which is an API that allows people to send images to Google, Google then gets information about it from its own AI processing, and then sends it back. A great example of this would be one of our clients censhare. They have a digital asset management system. And they will automatically add more information, more tagging, more data about an image by using the Google Cloud Vision API. So you don't have to worry so much about categorising your products or looking for colours or anything like that. It's all done automatically. And then if you want to have a picture of a blue ski boots, it's then very easy to find one in your digital asset management system. So it's a great tool. censhare, you know, is one of the companies out there doing this, there are a number of others as well. But I strongly recommend people take a look at the white paper that's on the show censhare website that talks about AI and machine learning if they're interested in learning more about AI and content management.

So we've seen some generic applications that help marketers, but ultimately, you know, we want the utopia of like some robot sat there typing into a machine to create our campaigns for us. So are there actually AI applications that really optimise marketing activities rather than more generic ones? Well, yes, there are. And so the first thing to say is that, if we look at account selection, ABM, there's a number of AI systems that really aim to help you find and target the best accounts.

So once you've got an account list, you can use a tool like bombora, this will go out and try and find intent data. So look on the public web, to find information about your target clients. And it will try and identify things that drive sales. So that could be for example, you know, if they hire new people, if they announced new sales, if they perhaps are a startup, they're getting new funding round, all of these things can can be indications of likely intent. And bombora will also look at things people post on social media as well. So you can go a little bit deeper than the very obvious things and look at, you know, what people are posting on whether the sentiment in posts, for example, is a good indicator to a company becoming likely to become a customer.

And there's automated account discovery. So this might be someone like Terminus, where you provide a list of accounts that you want to target, and the system then identify similar accounts. And there's a whole range of different tools that will do this. And depending on how complex you want it to be, you can obviously pick a you know, tool from very simple, you know, simple SIC code type analysis, I'll give you companies with the same SIC code in the same region with the same number of people all the way through to much more complex matching that's available with some of the better tools.

And finally, there's digital behaviour analysis. So and that's really taking, you know, what bombora is doing and looking at some of the contacts you're targeting, and really trying to build a picture of the company and the contacts and it's really trying to look in depth at whether, you know when would be the right time to approach that particular customer. So, the key key suppliers really in this are people like bombora and Terminus, as I mentioned, in fact, Terminus, I think has a deal to take boobers intent data. So it's a relatively small market. And it seems like there are companies really specialising in certain areas, and then partnering to get expertise and others. But this is I think, an area which is going to continue to grow as we look forward to mapping the customer journey is is is really interesting, this is a graphic on the right from a company called path factory.

And they talk about, you know, people coming to your website, first thing they get is a very long form, you have to fill it in, you then get a content asset, that content assets typically not personalised. you hoped they read it, you've then got to get sales to reach out and contact and, you know, clearly past factories view is ultimately this is not a very effective way to do things.

It certainly can work and I know a lot of companies that make it work, but it's it's hard work. So what power factory tries to do is dynamically serve content so that you know, their vision is the right content at the right time. It moves you away from serving PDFs into serving HTML. So you can track things like the time spent engaging on each part of the content, and link that to the ultimate outcome sale or no sale. So you can actually get a lot more information than offering a PDF. And, to me, one of the interesting things is a lot of people don't realise when you offer a PDF, and someone signs up for it, the only thing you know, is that the title of the PDF interested them, you really don't understand whether the content resonated where they found it useful.

Whether even you know, the content related to the the title that you gave the to the piece. So actually, you know, looking at PDFs, you get much less information. And talking to some of our clients, they've they've looked at this, and they've also found that, you know, it's really important to understand what people are looking at a simple datasheet download doesn't necessarily signify interest, there could be lots of reasons that people are downloading data sheets, you know, for example, if you think about semiconductors, people might download the datasheet, because they're doing a PCB layout, not because they're designing a product. And so with data sheets, it's actually really important to consider what part of the datasheet people are reading, in order to get an understanding of how likely they are to buy. And the only way you can do that is break your PDF datasheet up into multiple HTML pages. So this approach is starting to gain interest. And we're starting to see more and more people do it. One of the challenges, of course, is that it's relatively easy to pop up a form and have a PDF behind it, it's much more complex to split that PDF into multiple HTML pages, and then find the right time to gather contact details. So it becomes a much more complex thing to think about. And that's where companies like powerfactory come in, is they try and remove some of that complexity by automating the process. Of course, one of the issues is, is that you're going to have to have a fairly high level of traffic in order to gather data. So path factory can actually serve what's likely to be the right content, it can learn what people are interested in, and serve the content that they're likely to be interested in, content creation is, is another area and actually AI content creation is already proven. So we've seen things from baseball reports to stories about companies, financial reports, all being generated by AI's.

And there's lots and lots of AI content companies. So pasado, you know, is aiming to find out which phrases resonate, that's one of their key key claims. So they're looking to find the phrases that work for your audience. So you can then make sure you use the phrases that work the best phrase, he aims to do more in terms of actually, you know, pure AI powered copywriting. And then you also have products that actually identify content. So rasa.io is a very simple tool. And all it's looking to do is to find related content for newsletters, so you serve it, maybe two or three stories that you've written for your brand. And it finds similar stories to create a larger newsletter that focuses more broadly across the industry.

So it's it's already here, content creation, but there are many, many pitfalls. So the Amazon launch in Sweden is perhaps the the best known of these. And there has been some speculation that actually it was so bad that Amazon maybe did it deliberately to get PR but I'm not sure that's the case, what they did was they did automatic trans translation. And unfortunately, it led to some really bad translations. That might have been confusing, they might have made no sense. Or in the case of this T shirt, with an unfortunate translation of policy, it can occasionally end up with a vulgar product listing.

So it does show that, you know, even Amazon with their power and their resources, their automatic translation that had huge problems. So generating natural language is not as easy as you might hope. But there's a lot of people working on it. And it'll be interesting to see what happens moving forward, as to how quickly we can get AI generated content. To be honest, I think, you know, in the foreseeable future, that AI content will be fairly limited. So if you look at what's happening today, it's producing in natural language text, but from a very fixed input. So in terms of baseball reports, it will just simply talk about, you know, people who scored how many outs they were each, you know, each stage and things like that. So I really don't know about baseball, so I'm probably not the best person to talk about this. But the baseball score does tell you pretty much exactly what happened in terms of the major highlights. So they're just pulling data out of the baseball's score, and then putting it into natural language, to create content without that input with no structure is much, much more difficult. And there again, I think our copywriters can, you know, sit back and relax, because it's unlikely we're going to see copywriters put out of business, when it comes to, you know, writing, you know, real, genuine, innovative copy, maybe product descriptions could be written with AI in the very near future. But I think typically, you know, the kind of long form copywriting that's still going to be written by humans for the foreseeable future prospect engagement is a is a fascinating one.

And typically, this is around follow up emails. So what happens is you have someone download content, or you meet someone at a trade show, and then having to go and converse with them is really painful. If you do it manually, it's very time consuming, you can go to a market automation platform and create an automation. But again, that's quite time consuming. So now there's emerging a number of AI assistants that aim to do this to do the follow up. So they'll send emails that appear to be from real people, following up trying to get someone to respond.

And if you're interested in this, I mean, the great news is, is that you can start a free trial with products like Converse occur, Converse occur.com. And go on to the website, and simply upload contacts, and it will learn for a small data set will actually engage them and follow them up. Now don't get too excited. I mean, the interaction, particularly with a free trial of Converse occur is pretty limited. But if you just want something that's going to automatically send follow up emails, it's actually not a bad solution. So again, you know, taking a bit of the drudgery out of that follow up work by having these automated follow up automations is a really good thing

The next area is understanding engagement. And I think this is a really interesting area because fundamentally what it's doing is it's applying another layer of intelligence over your analytics data. And there's increasingly tools that offer content insights. So it looks at how and when people engage with the content, and ultimately aims to provide smart personalization. So not just personalising based upon the persona, but also maybe personalising, based upon the stage of customer journey that the system believes the website visitors at, based on what else they've looked at.

There's a number of platforms in this area. So I mean, Salesforce recently acquired engage. But also there's other platforms like dynamic yield, monetate, and platforms like that, that are all looking at how people engage with data and trying to build up potential customer journeys. by analysing the analytics of your website. It's a really interesting area. And I think this is a an area that definitely we're going to see some real benefits from in the near future. Again, it's not necessarily generating the content that people are going to engage with, but it's certainly going to help you serve the right information at the right time to visitors. And that is something that we're very close to at the moment and I think these tools are becoming you, almost the point where they can actually dynamically serve the right information. So it's an area that I would certainly watch very closely.

Email optimization is another area that people see, you know, you often hear these rules of thumb for email. So you know, you'll hear the send the email eight in the morning, or send the email just after lunch or don't send it on a Monday or anything else. But actually really optimising email frequency and time is hard, not least, because different countries will have different cultures, and the different parts of countries will have different cultures. So typically, people in the West Coast tend to start work earlier than people on the east coast in America, apart from computer programmers, who tend to start at about 11 o'clock on the west coast and completely mess things up. It's just not that simple. It's not that everybody has the same starting time. And if you want to be top of an inbox, you just need to send up one minute tonight, everyone turns their computer on, that is just not how things work. So even sending it local times is well away from optimising the email send time.

The other thing is frequency. More emails don't necessarily mean more OPT outs, we had a project a while back for a client,where we went from sending a monthly newsletter sending a newsletter every two weeks. And I have to be honest, we were kind of worried that doubling the frequency of the newsletter could actually result in more opt out. So we looked at it very carefully. The reality couldn't have been further from what we expected, though, we actually saw fewer OPT outs per month when we had double the email newsletters than we did when we had the original number. So less than half per newsletter.

So, again, you know, sometimes people actually they prefer to see things more frequently. It's really difficult to optimise this and it requires basically a machine to sit there and look at what works and what doesn't. And so we're seeing a number of tools and seven senses a well known product that aims to detect and act on engagement with email. So it's looking at things like opens, but also looking at whether those emails lead to conversions, and creases. Another important thing is people opening our emails are not necessarily representative of conversions for the email. So it's really important to try and look through that whole customer journey to see what works best when you're sending emails. And some of these tools are now able to do that.

So we've talked a lot about different products that use AI. But what about creating your own? Well, actually, it's really easy. You know, you can go to a service like Google TensorFlow. And if you've got custom data, let's say for example, you've got a custom database for design registrations, where you register every design, track it through and see if it converts, you want to find out what are what are the primary factors, that mean that one particular project is more likely to convert than another, you might need to build your own AI because there might be nothing off the shelf.

And you probably only need a relatively small number of lines of code to do that sort of analysis. So to go from, say, you know, registering a design to wins and looking at the various factors on the designs and how they impact whether or not you win that project. The problem is, although you don't need very many lines of code, you need a data scientist to write them. And you probably need a lot of data as well to do that. But we know I mean, but one client I talked to, he'd worked on a project a while ago where he looked at design registrations to try and understand, you know what meant that design registration was likely to turn into a win.

And they came out with two factors with which, you know, initially may seem a little counterintuitive, because you know, lots of people focus on pricing, and lots of people focus on registering new products and focusing on winning designs for new products. And they actually found that the newer the product, the less likely they were to win the design, and the more focus on pricing. So the more aggressive the pricing they provided, the less likely they were to win. But if you think about it, it is actually pretty obvious because people have familiarity and experience with older products. So it's much easier for them to design in. And furthermore, when you look at pricing, if the company you're working with, so the customer is not worried about pricing, it probably says that you have something unique that they really need. Whereas if they're really really concerned about pricing, then probably there are other competitors who have similar products and so it's going to be harder to win. So unless you're the price leader, probably you know low price might be an indicator that you're less likely to win a design. So it's certainly possible to create these applications. It's very complex, it does need a data scientist, there are actually now freelance data scientists around who will build your models. But you really need to have some level of expertise to know whether you're training the model with sufficient data and things like that. But there's certainly opportunities to build projects around creating AI that is designed specifically for your needs or your databases.

So we've talked a little bit about products available today. And we focus a lot on the fact that there tends to be, you know, fairly straightforward applications of AI. And you can create your own AI, as we've just said, but it's incredibly complicated. But what's the future? Well, in the short term AI isn't taking our jobs. I mean, that's the good news for all the marketeers listening to the webinar is that we've still got, you know, quite a bit of time left, where AI is not going to be able to do everything we do.

You know, ai often has limited capabilities. Sometimes there's very small datasets, that's impossible to train an AI. And so you might have to draw analogies, something a human can do something an AI today finds very difficult. And often there's no data at all, which clearly is very tricky for an AI.

Having said that, though, AI is going to help us more. And I think, you know, if we look at where people should really be focusing on AI, I certainly think the generic tools, so image and voice recognition, natural language processing for chat bots, all of these tools are now quite mature, and are really ready for you. So if you're not using those technologies for your digital asset management, or to create chat bots, now's the time to start thinking about it. Campaign optimization is another area where, you know, definitely I see people getting benefits. We talked a lot about, you know, the understanding of the journey by looking at what content people view. And by splitting out PDFs into multiple HTML pages. And this campaign optimization, I think, is an area that really is about to hit primetime, it's a real important area to look at. And you might not want to deploy it today, it's still an expensive technology. But in the next few years, I think the costs are going to come down. And we're going to see a lot more people using AI and campaign optimization, particularly in terms of dynamically serving content on the website, to drive people through their customer journey. And finally, performance insights. And I think AI's are going to be able to give indications on performance, whether that be, you know, whether you're likely to win a particular design opportunity, or whether your lead is a high score, or a low score or any of these things. So I think AI is going to help us more and more and taking advantage in these three areas. I think they're the three areas where you're most likely to see benefit in the near future. But just like our client, ABB that talks about cobots, rather than robots, and the foreseeable future, is that that tool is going to be helping not replacing us. And I think you know that the same is true to a large extent in robotics. In most manufacturing environments is acumen and robotics environment. And I think, you know, if we look at marketing, it's going to be an AI and human environment, as well for you know, certainly the next few years.

So finally, how can you take advice, take advantage of AI? You know, we like our top tips at Napier. So, here's our five top tips. And one bonus tip. So the first thing is don't feel we're being left behind. Although people are talking about AI, the actual use of AI is pretty straightforward. So you can get on board very, very quickly without needing to invest a lot in terms of data scientists. And you do need to understand the difference between different AI's between using formulas and algorithms at one end, and machine learning at the other.

I would certainly experiment with simple AI, probably the easiest way to do this is with Google ads. You know, my experience with Google Ads is sometimes the AI stuff is is awesome. It's absolutely brilliant. And you can't get close to it manually. Other times you look at it and you say well, why on earth is that doing so? You know, sometimes we see great results, sometimes not so but certainly experimenting with AI and try to understand how to get the best out of it is really important today.

Keeping up to date is very important. So certainly follow what's going on and try and understand who the new vendors are in the market. Build your data sets. We talked a lot about the need for for large amounts of data to train, machine learning models. And so build your datasets the more data you can build now, the more you'll be able to use AI as those machines learning Tools come on board.

And then finally, and here's our bonus tip. And I'll admit it's entirely self serving. But your agency should understand I should be talking about AI, your agency shouldn't be saying AI is going to replace everything that they do, because that simply isn't the case today. But an AI, an agency that uses AI is going to be a more efficient agency.

So the long and short is, you know, although few organisations are really using heavyweight marketing AI, there are real applications that can be delivered with, you know, very, very small, very reasonable budgets. So it's important not to get left behind to get on board now and start understanding how AI can benefit you.

So that's our overview of AI and b2b marketing. I think it's it's a really optimistic view. You know, already today, there are ways you can make use of simple AI to help you in your work. And looking forward, I think AI is going to become more and more of a benefit to us and more of an assistant. But at the same time, you know, there's no indication that we're all going to be put out of a job. And everyone's going to see the same kind of AI created campaigns for every company, that's just not going to happen in the foreseeable future. So I think it's a very bright future for people are prepared to engage, and to frankly, try a few things. And so now we'll move on and see if anybody has any questions

by Hannah Wehrly