Discover how to use AI to create email marketing campaigns with Napier’s on-demand webinar.

The webinar covers:

• Using AI for contact research
• How to integrate AI into your Martech stack
• How to write great emails with AI
• Testing with AI
• Limitations, and risks with an AI email campaign

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.

How to Use AI to Create an Email Marketing Campaign

Speaker: Mike Maynard

Hi everyone. Just to let you know that we will start the webinar in a couple of minutes. If you want to get yourself ready, maybe get a cup of coffee. And as I say, we’ll get started around about 3o’clock UK time. Okay. Hi everyone. Thank you for joining This webinar, we’re going to get started. We’re going to talk a little bit about AI for email marketing. So, what I wanted to do was with this webinar is really dig into some of the realities around AI and understand whether AI can deliver great email marketing or not, so we hear a lot of people talking about the importance of AI, and I think most people, in some ways or another, are using some sort of artificial intelligence in their email campaigns. But what we’re going to do is we’re really going to dig into it, and we’re going to find out you know what works and what doesn’t through some practical tests. Now, obviously, these tests are somewhat simplified, but hopefully, it’ll let you see some of the challenges that could arise when you’re trying to apply AI to webinars to email, and also will hopefully give you some tips as to how you can overcome them. So, the first question is really around, you know, how can you use AI to help your email marketing campaigns? And I think the important thing is it’s not just writing emails. So today we’re going to look at a variety of things. We’re going to look like at segmentation. We’re going to look at timing, and as I call it, other boring stuff-things that perhaps aren’t so interesting, but maybe matter to some people. We’re going to look at AI for idea generation. Then we’re going to look at AI for reviewing, optimizing, and testing your email campaigns before you actually launch them. We’re going to talk about personalization. I think this is probably one of the things a lot of people are quite interested in looking at: is how you can use AI to personalize or even do recommendations when you send emails that are designed specifically for each recipient. So personalization at scale. We’ll have a quick word about some of the tools and techniques, and lastly, we’ll give some recommendations. So let’s kick off with segmentation. Now, this is kind of interesting because I think most people in B 2b tech recognize that things like sit codes, the existing categories for industries. Are really way too broad. They’re not helpful because they’re mainly not specific enough. And so, one of the things that’s really cool about AI is AI lets you have much more subtle segmentation. And not only that, it lets you control the market definitions. And so, you know, my example would be if you look at the, you know, say the mining market generally, people would have very different definitions of that depending upon what they’re looking to sell into the mining market. And the great thing about AI is when you ask AI to do some segmentation, it’s really good at following your requirements. But the one thing to say is that AI segmenting data is not necessarily a simple, straightforward process. It’s normally very much an iterative process, and we work on segmentation using AI. We’re going back and forth several times to optimize the the segmentation. So it’s very much an iterative process. You use AI first to learn about your data, and then you can start using it to segment. And normally that takes two or three goes. So, just a warning: what we’re going to show here is perhaps a little bit simple and jumping a few steps, but it should give you an idea of what’s possible. So let’s have a look at this. I’ve pulled this from a recent census show in America. So at the census show, there were a whole bunch of companies that were all classified by the census show as semiconductor manufacturing. Now, when I look at some of these companies, there’s small companies, there’s large companies, there’s companies, you know, making products that are processing sensor data, companies making communication products to trans transmit the data. They’re all very, very different. Semiconductor manufacturing is just way too broad, and it’s a great example of how we often see our data. It’s actually true. We see that data quite often in first-party data. So where you know clients have been gathering data.

They may have asked for an industry, but generally speaking, the industries have to be fairly broad because you need a fairly small number of choices in order to make sure you get good data and high completion rates on forms. So I’m stuck with this. What do I do? Well, the first thing I can do is I can actually go and look at what industry AI thinks, so we’ll just put in a prompt that basically asks what industry does each company, and also provide the web address to the AI to let the AI go find the website, actually fall into. And this is what we get out. Now you can see this is much more specific. Suddenly we’re straight into some very, very specific things. So, whether that be you know photonic semiconductors, analogue products, microcontrollers, sensors, whatever it is, we’re now getting much data, but much more data. Really insightful stuff as to what the company does. Of course, this is AI, so the standard caveat applies. It’s not going to be perfect, and I think the reality is with a lot of campaigns where you’re looking at doing these bulk mailings, you’re going to accept that AI is going to make mistakes, and you’re going to be happy if it gets you know 90, 95% of the industries right. So now we’ve got some really you know in-depth information. This isn’t super helpful, though, because if I want to start, you know, for example, sending different emails to companies that use MCUs versus produce MCUs versus companies that produce analogue products, it becomes quite complicated because we’ve got quite comprehensive descriptions. So, what we can do is we can actually ask the AI to narrow down to a list of different industries, and this is exactly what we did with the data. And you can see here here that we produced a few categories. So, we told the AI to categorize it either as analogue, wireless and communications, microcontrollers, video processing, or other, and you can see we’ve got a couple of others that in there. So now this becomes incredibly powerful. So you know, in a couple of jumps, we’ve now got to the point where I can send you know analogue-specific emails to companies who make analogue products, wireless to those that do wireless and communications, and so on. And obviously, it’s an example of you know companies that are potential clients for Napier. But exactly the same thing will apply to you guys when you’re looking to sell your products and services to potential customers. Using AI to segment is really effective, and it means you can have pretty personalized emails based upon rules rather than necessarily based upon creating those emails straight away. So I can have an email in this case for companies that make analogue products, companies that make wireless, and it feels much more personal. We’re getting much more specific, and that’s going to increase the response. Rate of people who receive emails, so it’s a way to very quickly improve your performance. AI is great at segmentation; it’s something I’d highly recommend doing, and it’s obviously something we do a lot for clients. The one thing I would say is this is quite a straightforward example. Sometimes digging in and deciding those categories is really the key thing. How you decide to split up the companies, so that does require a bit of thought, and that’s something that, to be honest, AI has not been great at in terms of recommending the categories. It’s not bad as a first pass, but it’s never something we find works in terms of a really good set of categories. So we’ve now segmented it. Let’s have a look at getting some contacts. I’d love to get some contacts. The problem is, is AI kind of sucks at researching contacts. So if you try and then within your spreadsheet that you’ve got, try and ask the AIs to go out and pull in contacts. Generally speaking, AIs are terrible, but there is some good news. If you have contact databases, so for example Apollo, it has an MCP server that allows you to interface an AI to the Apollo database and pull contacts out.

Also, there are products like Clay that you may be familiar with that actually designed to pull contacts out and tie in AI with contact extraction, so AI kind of sucks on the contact side, I’m afraid, but there are tools that you can use to add in to get those contacts. Another way we can segment is by looking at the intent or where people are on the customer journey, so AI offers this potential to really segment and either segment based upon what the person’s done, so what we think they want, or what the contact’s done, so where we think they are in the customer journey. Do we think you know that they’re down towards that decision-making stage, looking at data sheets, or are they looking at much more high level, trying to you know source vendors, and earlier on in the customer journey, or maybe even doing just research as to possible solutions. And so AI is actually pretty good at this. There is a challenge, though. Whilst AI can help in designing that, in in understanding where people are in the customer journey, actually most B 2b companies don’t have a lot of data, so the AI really struggles to learn unless you’ve got you know 10s of 1000s of contacts that you can train on. What actually happens is that we find quite often defining certain activities as being an indication of you know either purchase intent or you know early stage customer journey or whatever you want to use, that often is a much better way of doing it than trying to have AI, just because the data is so limited in a lot of situations. And one thing to mention though is whilst AI kind of struggles in terms of positioning people unless you’ve got a lot of contact data. It can help a lot in designing the customer journey. So you can pull data that you’ve got from what customers do and use that to actually you know create a better model of what the processes that customers go through from you know initial research through to purchase, so AI can help there a lot, but we’ve found it really struggles to actually rank people in terms of position unless you’ve got huge volumes of data. You know, some of our clients, you know, have very very small volumes. I mean, I remember, you know, working with a client that made baggage handling systems for airports, and there was literally one customer a year. So, forget about using AI for you know customer journey analysis. There, you actually know where that customer is because you know when that airport or the airport terminal is going to get built. We do find AI is built a lot into market automation tools for identifying intent and also lead scoring increasingly. It’s not a bad thing. I would say that you know typically again we often find rules-based scoring works better than AI-based scoring because there’s often some rules that are very clear and very straightforward. So you know you know who are going to be your best, strongest customers, you know those companies. You know you know how they’re categorized. You know your ideal customer profile, and if you don’t, you certainly should do. So again, rule-based stuff can often beat AI here when we test it. One thing AI is pretty good at is optimizing the sending time and the frequency of sending, so whether that’s the day you send, the time of day, you know, capping the number of emails you send, things like that, it’s great. You know, with most email tools, you can just put that on autopilot and let it run. Again. It does require volume, and so if you’re not really proactive in your email marketing, you may not be getting the volume to really get good data. And the big problem with AI, as we always know, is that if you ask it to make a decision, it will very confidently make a decision whether or not it’s right or wrong. So one thing I would urge you to is, you know, really make sure you’re feeding AI sufficient data. We do find as well that some of these optimizations have relatively little impact in terms of performance.

You know, picking the day of the week that you send your email on might only increase you know email open rates by perhaps 1% So it’s fairly small. If you’re sending high volumes again, that really matters because that can be a large number of recipients opening the email because you send on a Wednesday rather than a Friday. But you know it’s not always as big an impact as you think. So again, think a little bit about volume. AI loves volume, and so you know really think about whether you’ve got sufficient volume to feed it the data, so AI could do quite a lot in terms of the you know the the process around the mechanics of the email, who you send it to, when you send it, etc. It’s also obviously great at generating ideas. However, I would say one of the issues is, and we see this particularly as emails get more and more technical, is emails become more and more formulaic. And you could actually go out and rather than use AI, you could simply go and Google, you know, best B 2b email marketing template or best technical marketing email marketing template, and you can get very similar results. Not surprisingly, because actually, that to a large extent is what the LLMs are doing. They’re going out, they’re pulling data on the web, and then regurgitating it to you. So, I do think you’ve got to be very careful about being formulaic, and obviously make sure you edit the copy. So typically, again, this is an iterative process. You know, here’s an example. I want to draft a marketing email to engineering managers. I’ve got this made-up company called Sensor Magic that makes temperature sensors, and I’ve asked it to highlight three reasons. Now, we deliberately put a prompt in that was somewhat vague. Hopefully, if you’re prompting AI, you’re going to put a much more detailed prompt in to get an email back. But this is going to kind of rely on AI producing what it thinks are the benefits of the product. So, here’s the first draft of the email, it’s quite long. I would say that the claims are fairly vague. Greater measurement accuracy is not really something an engineer is going to look at and go. That’s definitely going to help my product because it doesn’t tell you how accurate. It doesn’t give you relative performance. You don’t know whether it’s better or worse than competitors. So I think it’s important to realize this is fine, you know it’s okay, but it’s certainly not a great email as the first one. It even offers to make this more technical, more punchy, or more focused on conversions. So what we did was we actually took that suggestion and told the AI to make it more punchy. We also gave it some more information about the centre itself. So, as you see, you know the the AI generated an email talking about cold chain applications. So, basically, measuring temperature in refrigerated and frozen transportation and storage. If it’s a room temperature sensor, not appropriate for that. So you know it’s kind of to illustrate that AI is going to make stuff up. So again, you really have to check carefully. And we’re also trying to say you know be really specific around accuracy. So let’s go back again and see what happens, and here I think you know we’ve got something that’s a little bit better. You know, it’s a little bit clearer, it’s a little bit shorter, a little bit more punchy, and it does try and highlight benefits. You know, both of the product and also ultimately with those three points of what it can do for the end user, so the equipment that the customer is going to make. We then had another go, being more specific about the accuracy, and again we get something that’s a bit more technical. So we told it the accuracy of 0.1 degrees C, and it gives us something that’s a little bit more technical, a little bit more specific. And then finally, we told it to be more technical, and we told it to halve the length. And we’re getting to something I think that you know is looking quite good. It is fairly specific now. Whether you’d want to lead on, you know, things like tighter control loop input, probably not.

It’s not necessarily a great thing to tell an engineer. They care about the performance of the product. They can work out their control loop is going to be better if the input’s more accurate. But I think you know you can see back and forth. We’re getting some quite good ideas from AI that we can then put into a final email. But sometimes it’s better to be more focused. So rather than say write an email, let’s look at five different subject lines. And here we get something out that I think is you know perhaps even more interesting. It’s producing different ideas that gives you different approaches to promoting the temperature sensor product. So you know we’ve got anything from you know a very factual 0.1 degree C accuracy to a question to you know kind of challenging statement. You know when one degree C accuracy isn’t good enough, and that’s really useful. It’s really interesting to be able to do that. So, I would recommend rather than just trying to tell AI write my email, sort sort it out for me, get it to give you options, get it to present different options based on different premises. So whether that’s you want something super technical or you want something you know more salesy and punchy, and then look at how those emails feel to you. That then gives you a really good idea. But once you’ve got that email, you’ve come in, you’ve maybe used AI to help you, you’ve generated an email. AI can actually be quite helpful. So AI is is great for reviewing and editing, and it’s also really useful when you want to optimize and test. So we’ve talked about this in previous webinars. Synthetic personas are really helpful with AI. So build your own GPT that’s synthetic persona, base it on your ICPs, and then use that ICP to really go and analyse the potential options. So let’s look at what we did. So basically, we built our own GPT, so custom GPT, around an engineer that develops IoT systems that use temperature sensors. The engineer is super technical, hates marketing, and is very cynical, rather like many of the engineers you’ve probably met. And then we’re going to ask it to suggest some subject lines. And so we get four specific subject lines here, where we can actually use it. Use different things. We also get some input on whether percentage or absolute accuracy is more important, and some comments about you know the words to use and words not to use. Now, to be clear, this is one ICP that you’re targeting in one custom GPT. If you want to do this in a real application, we would typically have multiple custom GPTs, so a couple of different versions of each ICP, so we can see what the impact is. Because obviously, you know, you’re going to reach different people with your email campaign almost certainly, and so you want to see how different people react, and then you need to aggregate it together. So again, a simplistic thing, but you can see how it can help you, you know, understand what to do. One comment I would say is that you know, even for a very technical engineer, these subject lines feel to me to be quite long. I’m. You can also take these and throw them back. So we’ve got the subject lines that were produced earlier, and we’ve just thrown them into the custom GPT. Those are produced from the general LLM, and we’re putting them straight back into the custom GPT. And we’re saying, you know, you as a technical, cynical engineer, what would you like and what would you dislike? And again, this is great. So we’re getting some specific feedback here. I do think you know there’s an opportunity to to look at what this says, and I think the most important thing with using AI to optimize and test is that you need to get the explanations. So here we’ve got a great explanation that node-to-node variation is great because it sounds like a real engineering problem. Of course, if you’re, for example, building standalone air conditioners, you actually don’t care about that because there’s only one temperature sensor, so there’s no different nodes to worry about. So again, you’ve got to be really careful, and you’ve got to apply human knowledge to understand what actually would work.

But it gives you some really good input, you know, and it also lets you take what you’ve got already and maybe build on it. So again, we wouldn’t simply throw some. Headlines into a custom GPT, pick the one that’s ranked the best and send it. Actually, what we do again is iteration. Work with the GPT to make a better headline rather than just trying to pick one from five. So, AI can be really useful in terms of testing and optimization. Let’s look at personalization. Personalization is really interesting. Generally speaking, personalized emails are going to be much more effective than un-personalised emails. You know, there’s no end of test to do that, and I think common sense also means that most of us, you know, intuitively understand that. There’s another benefit as well for personalization that I think is super important, and that is if you’re changing the email text you you send, you’re actually more likely to avoid spam filters. Obviously, as long as we’re not generating spammy content, but if you send a large number of the same email to one organization, or you know, you send a large number of emails to to multiple people at different organizations that can get flagged because the email filters see the same email over and over again, and to them they see it as spam. So it feels good. It feels like it’s the the first thing we should do is absolutely look at how we can use AI for personalization, and what we see a lot in the consumer sector and the less technical sectors of B 2b is that AI is becoming quite widely used in terms of personalization. What we see with some of the more technical markets is that it becomes quite difficult, and I’m going to be honest. It kind of looks like AI tries too hard. So what we generally find is that AI can be, you know, too sycophantic, a little bit too nice to the customer. It can also get some of the technical details wrong when it tries to go off and write something a little bit different, and so this is what we did. We we basically briefed an AI to produce some some text for different recipients. We picked you know four companies that could potentially buy this fake company’s Sensor Magic’s product, and we got it to create emails to individuals who are real individuals, these are real LinkedIn profiles, who actually work for these companies and are actually involved in IoT or temperature censoring design. And you can see here the text is not bad. I mean, I’m not sure I’d do first name surname as as a intro, so you know the first one kind of makes you cringe already. You know it’s also here, and I can tell you with ABB, it’s actually what it’s done is it’s pulled out information from the ABB website that isn’t relevant to IoT temperature sensors. So it’s pulled things like fast load changes, which really apply to ABB’s electrification division, and doesn’t apply to temperature sensing. So we’ve got some technical things here that, if an engineer receives this, they look at it and go, “That’s kind of wrong. Most of it’s pretty good, but there’s a little bit there that kind of feels a bit, you know, inaccurate, and you look at that, and that is almost always a you know immediate flag to say this is is almost certainly AI generated. I think some of you will probably know that at Napier we’ve got a couple of podcasts, so we’ve got marketing B 2b technology, and we’ve got the marketing automation moment. Marketing B 2b technology is is a guest based podcast, and it’s really interesting because about a year ago, I got a couple of emails where people email me, and they they listened to one of the episodes, and they really liked something from the episode, and they wanted to be a guest, and and I felt really good for these first two emails, and then I got another that wasn’t quite so well written, and felt like someone hadn’t really listened to the episode. And I very rapidly realized that a lot of people were using AI to pitch for podcasts. And once you see it, it becomes very, very obvious that the text is AI generated, and now I can absolutely pick.

You know, probably I would say somewhere around about 80% of the pitches we get for guests for our podcast are AI generated, and that’s okay because it doesn’t necessarily mean the guest is a bad guest, but it does mean that there’s all this effort around saying that the guest has listened to episode, you know, 400. They haven’t. I know they haven’t. It kind of doesn’t, you know, build that relationship as well as you should. So I would say the most important thing that the one thing we’ve learned with generating content for emails that’s personalized by AI. Is put really really strict guardrails on, and that really is the key thing. If you can get away from more general stuff, so I mean this is an example that I talked about here. This is another one that was produced as well. So just to make it easier to to read and see, and you can see here that you know some of this is is not bad. I mean, this is not a terrible email. It just doesn’t feel to me like an engineer’s written it, and it actually doesn’t really feel to me like a marketer’s written it either. So, I think it’s about taking this and understanding how you can apply guardrails. So, our challenges really the biggest issues are errors, and we’ve seen errors. I mean, my my favourite error actually for AI personalization is around missing out the word module. So we had a campaign, and we were trying some AI personalization, and the AI put image sensor instead of image sensor module. And of course, the problem is is that actually different people will decide on a module versus a sensor. So if someone’s going to buy in a a sensor, they need to put it on a PCB. They’re an engineer. Whereas if someone’s buying a module, they could well be a systems engineer. They’re they’re not an electronics design engineer. So errors and particularly very subtle errors around technical phrasing are really important. As I say, the language used is is an issue as well. The AI content it doesn’t really sound like an engineer a lot of the time, and also AI can sometimes not understand the recipient, and that’s because we struggle to give AI really good details about who the person is we’re sending it to, and that, to a large extent, is because in you know B 2b tech, a lot of the databases don’t have a lot of detail and colour about the recipient, particularly in terms of things like preferences or personality or anything like that, the kind of marketing material they like, and so you can end up sending a you know a fairly casual email that works very well in the UK. You then have the same personalisation that goes out to Germany, and it really doesn’t hit because German culture is very different, so understanding the recipient as well is a is a real problem. I think it is a problem that it’s easy to spot AI emails. As I said before, you know, with our podcast pitching, you know, I now spot them and I’m happy to look at them and read them. But it still grates a little bit to feel that they haven’t bothered to actually write, you know, an email. They’ve just used AI, so I think working too much to get too much AI personalization can actually be counterproductive. So as I say, you know, the one thing we’d say is keep AI within tight guardrails. And often when we’re using AI for personalization, what we’re doing is we’re narrowing down on particular phrases and getting AI to write particular phrases, not even a sentence, and certainly not the whole email, because that’s the way to really make sure it doesn’t go too much away from what you want to see. And then the last thing to say is, particularly with the benefits of AI segmenting data really accurately, automation is often a better way to go about it than AI. So, rather than use AI to write the email, we quite often find because you get much more control and you can get that language precise. If you’ve got something super technical, it’s better to substitute certain fixed phases depending upon characteristics of the recipient.

You know, for example, industry or job title than it is to get AI to write it from scratch. So, don’t rule out automation. Automation is really good. So that’s really covered some of the things we’ve analysed for around AI and AI and email marketing. I’m sure people are going to be interested in the tools. So here are you know some of the tools we use. So obviously the large language models, ChatGPT, you know, and all the other models are really important. Just in case anyone’s interested, we use ChatGPT for this. I’m a big proponent of not being too hung up on a particular model. I think if your workflow relies on using a particular model, it’s kind of vulnerable. You know, if that model upgrades or changes, and you’re so reliant on a particular one, it could be a problem. Yes, I agree. There are better models and worse models, and arguably Claude is producing better written content than ChatGPT today, on average, but I’m sure that’s something that you know is going to change as we go through forward in the future. We make quite a lot of use of GPT for Sheets and GPT for Excel, so this is allowing you to insert queries into Excel spreadsheets and populate data based upon queries. Through the API, that’s really important. We also make use of Make and Zapier, so again, automation tools to go out and you know send prompts to the API of different LLMs and get data back and put it in. That that again is a tool we use a lot. I mentioned databases. You know, we use Apollo quite a bit, and Apollo has an MCP server that allows AI tools to talk to it. So that’s one way to pull in contact details. The other way is a tool called Clay, which kind of integrates the contact and the AI together. And lastly, I think you know if you’re looking for tools that make use of AI, marketing automation systems are always great tools, and we’re seeing more and more AI integrated into those systems. I would say you know always test and make sure you feel it’s right, particularly if your products are at the high tech end, because this is where we find LLMs have the problem. If you’re looking for something fairly general, you know, inviting someone to a trade show, you know, kind of my view is knock yourself out. Use AI. It’s not going to go very far off pieced, but even something fairly simple like a temperature sensor and trying to pull what companies do from their websites really can be a problem for AI. So the more technical it gets, the more careful I think you have to be of deploying that AI, even if it’s built into one of the systems you’ve got. So thank you very much. I’d really encourage you, if you’ve got any questions, to put them either into the chat or the Q and A. We’re actually going to take a little some holiday, a couple of months off from webinars. So we’ll be back in September. If there is anything you’d like to, you know, to see as the next webinar, please let me know. I couldn’t resist putting this in. You know, this robot’s been with us throughout the presentation. Obviously, AI drawn, and I’m not quite sure, but it seems to be sticking the the straw into its nose rather than its mouth whilst drinking the cocktail. So, again, a little bit of a small AI fail there on the the picture, just to indicate that whilst AI can do some great stuff, when the details matter. That’s when you’ve got to be careful. I think that’s you know something that a lot of us need to remember, because a lot of us are working in very technical industries with very technical products. So, as I said, I’d really invite you to ask some questions. So, if you’ve got anything you’d like to ask, as I say, please do drop it in the chat or put it in the Q and A. I do have one question already, and someone has asked about how we interface into different AI tools, and so that’s a really good question because a lot of what we show today has been done through the chat interface, and that works quite well if you’re doing a one-off.

But if you want to do something in volume, so bulk customization, personalization, it doesn’t work. And so really, it comes down to looking at how you can automate things, so one way you can do it is by using spreadsheets, and as I mentioned, the AI interfaces to both Google Sheets and Excel. There’s add-ons you can get that will then interface to the APIs. You create an API account, and you can do that. You can also use the middleware tools like Make and Zapier, and they’re great. And particularly, there I think you know something like Zapier is great for when you’re triggered. So perhaps if you you know get a new prospect or a contact filling in a lead form, you can then go out and ask AI questions about the company and things like that, and get some rich data. So I would say, you know, using that middleware. If anybody’s interested in learning more about the middleware, please do ask us. We do a lot of this both internally and also for clients, and so we’d be more than happy to help you. I know I’ve overrun a little bit longer than we normally do on these these calls, so I don’t want to keep people any longer than we need to. I don’t see another any other questions, so unless anyone has a question for me that they want asked, you know, straight away. What I would say is, you know, take a look. My contact details are there. Mike at napiev2b.com. If you’ve got any questions, would like to know more, or would like us to help you into to integrate AI with some of the email marketing systems you’re using. Please do let us know. If not, thank you very much. Have a wonderful summer. I hope you all have you know some great holidays, some lovely weather, and look forward to another webinar in September. Thank you very much.