-Hello, everyone. -We'll give everyone a few more seconds. Just join in and then we'll get -right into it. -All right. Okay, let's get started. Welcome to today's webinar. Today we're going to be looking into lead enrichment in monday.com using AI. Big part of that is as always with my webinars Make. So let's get into it quickly mentioning the agenda that we're going to look at today. First off, we're going to quickly go through some practical AI applications, just looking at what that actually means and how I can actually use in a practical setting. We're going to go through the use case that we'll be looking at today, kind of go into some more details of what functionality are we looking to support? We're going to have a look at a conceptual design to just look at on a high level, how do we actually implement AI into another system like monday.com. And then we're going to have both a demo where we'll actually live look at how it can look as a user. And we're also going to do a walkthrough where we look at the back end and the actual functionality, how it works, how it's set up, and something that you will also be able to replicate after today. So starting off with a quick discussion about practical AI applications, if you're one of the people who have went out there and tried out some different AI concepts, some use cases, some functionality, there's a lot of stuff out there that is quite gimmicky, and some of it is really useful, and some of it can be very narrow down. Generally, what we're looking for when we're talking about practical AI, oftentimes it's a very automated function and a human check. What that means is we try not to do too much with the AI application. At the same time, there's things that should help us out in our daily work, but at each step, there's some human interaction. And really what we're looking to do is replicating dynamic human tasks, non dynamic human tasks, meaning something that is very repeatable by a computer. For example, every time I click a status, I change it from. This is something I'm not working on. For example, set a date somewhere else or move it somewhere else to another space. That's a very repeatable behavior and generally quite simple to automate, especially in monday.com. However, when you get into something where you actually have to look at something with more human eyes to make a decision, let's say, for example, a very classic. When you try to avoid duplicates in your accounts registry or client registry in your CRM, a lot of times it's hard to remove duplicates or notice duplicates because the same people or even different people might spell the company slightly different. And that way it's hard for a computer to tell them apart. Or even more important, tell that they're the same. Whilst for a human this can usually be a quite simple task, those are the kinds of automations where I can really come in and shine and automate those tasks as well. So going into the use case for lead enrichment, the use case that we have here is that we have leads coming in from multiple sources. It might be a contact us page. On our home page. It might be manual input. It might be a third partner supplying us with leads. But the starting information is quite basic. We might only have information like the company name. And in this case we're also going to be looking at the company website. What the use case is that each lead is enriched manually. And the tasks that has been difficult to automate is that for each lead, we're going to be going into the website, and we're going to try to figure out some key aspects of that potential client, and then we're going to go back to our leads and fill that in. And this is today done by a human that reason that it's very hard to automate using more standardized automation software is that, of course, if you go into a bunch of different websites, the website structure is going to vary a lot. Each website looks slightly different, some websites contain different pieces of information. And then on top of that, sometimes you can have different languages and etc. so this makes it very difficult to automate without using AI. Then we also added a small, small piece of information outside of that as well, which is that at the end of this lead qualification stage, we have a kind of semi called pitch by email. The kicker to this is that we don't always want to send the exact same standardized email. We want to make these standouts slightly more custom personified to that specific client. And we're going to look at how we're using AI to do that as well. So this is really roughly how it looks from a conceptual standpoint. It's not really that difficult to to follow through. We have a starting point here in monday.com. This is where we initially receive our leads. At some point we're going to send a signal to Make.com, which is going to utilize the ChatGPT API, and then do some magic with that in the background, and then simply throw that new information back into monday.com. So let's have a look at the demo. Okay, so here in monday.com we're currently sitting in the sales CRM product. We have a very barebones lead board. There's plenty of more stuff that can be added here, but for just looking at it today we've made it super simple. So right here I've just added in a very basic lead. We have a lead for Omnitas consulting. We'll go ahead and assign a responsible for it. And then the next step for us would be to fill in some information about this client. What's the market they operate in, what industry, what target audience do they target, and also what's their value proposition? Usually what we would do is we would go into their website, we would look around, scroll around and try to figure this out and then fill it in. But in this case, now we're simply just going to fill in the website and in the background now what's happening is the API is essentially doing or sorry, Make is essentially doing the same thing I just described. It's fetching that website. It's doing some magic to just kind of pull out the source information that is important, and then it's just giving ChatGPT a bunch of information, taking the output from that and then putting it back into monday.com. So here we can see now that it's put in some values so honored as consulting. Obviously being us were in the B2B market. We work in the software industry and our target audience is generally sea levels. Here we also have a value proposition, which is Omnitas consulting helps businesses streamline workflows and achieve -growth and success. -Great. But. So let's actually look at what that looks. Right. So this is the execution we just did. Let's go into it and take a quick look. So for those of you who haven't been in Make.com before, it's a no code or very visual way of programing. It's a way to connect different systems together. And on top of that, you can also implement a lot of your own logic, but most of it is always something up happens and then it goes through a chain and leads to something at the end. So in this case, let's go all the way to the left where it starts. We start with what's called a webhook. This webhook in monday.com is quite simple to implement. We have it right here with together with the automations we have when client website changes, send a webhook. So when I put in that client website that sends a signal to make and that's what this pops out as. After that, the next step we're going to get the website info. Obviously this isn't always the same web page, so we actually retrieve that information as well from this initial pulse. So looking into here we'll see this what's called a mapped value. When I mouse over it we can see the first module starting to pulse because that's where that value is retrieved from. And this is the URL. So this is simply the website address that we filled in. After that, we're going to do some other magic right here. This is something that's really in broad strokes, just means that the first piece of this is going to search for the header information of the page to just figure out what is this page about. Next part is very similar to that. It also tries to figure out the body content, so the paragraphs and everything. And then finally, this is simply just a big function to put that together so that it's ready to send off to ChatGPT. These can look very different and can be a bit more technical, but I won't get into too much detail about that part today. But if you like, you can try replicating that yourself. Getting into the chat GPT. So here's really where the. Good stuff is. First off, just some information. If you want to try this out yourself. In order to do this, you're going to have to register your own ChatGPT account. There's a free tier available and you can test without actually paying anything. But in this case, I've used a model that most free users will have access to. And then down here we filled in what's called messages. If you've ever filled in a or sorry, if you have gone to the ChatGPT page where you can actually sit and chat with the AI, this is more or less the same thing. So here the message that we've set is that this is the company, and then we've included the name, and this might be the URL. We've also said this is the content of the website, and this text is again from this. Parsed piece of information. And then after that, we're going to include a lot of other rules and specifications as well. So moving on we're going to say only respond with data, you know. And if you do not know it leave the value empty. We don't we would rather have no data than false data. And then we presented with the task, give the value proposition based on the content of the company in less than 30 words in English. Casual tone. There's a lot of text here. This can be really tweaked back and forth, and you can get this to do very specific things, or you can get some more broad things. In this case, we've actually supplied it with our list that we want for the different levels. So we supplied it with lists and then we let it make its own decision. For which one of these does it fit more closely into. And then we also asked for the other pieces. So market and industry. After that, we're going to do the same thing again. We're simply going to format that data to make it a bit simpler for us to work with. It just looks like that. And then finally we're going to add that back into monday.com. So in this case right here again we have mapped what's called a pulse ID or item ID from the starting module. This is simply just to say update the specific row that you just put in a website on. Since we don't want to update some other row. After that we put in the different column value. So we have market filled out in with market, industry with industry etc.. So these are the four columns we filled in. And it's really that picture that we looked at before. It's monday.com through Make and then back to monday.com, which is what we're seeing here from monday.com back to monday.com. I'll pause quickly for any questions. -Great. -Okay, so moving on then. That's step one. Uh, again, these are some of the pieces that we chose to display. You can have a bunch of other things in case it's something else that you want to include in your leads process. Moving on a bit. We're going to look at the second part of this. So now that we've pulled in some information automatically, we might supplement it with some other information as well. We're ready to create this customized pitch. In this case I've also added a manual entry. Let's say we've gotten this lead, we got some information about it. And we have a set of services that we sell. So for example taking Omnitas Consulting, we do some monday.com consultation and training and we do Make.com consultation and training. So in this case I've looked at this specific potential client, and I'm figuring I want to pitch for some monday.com consultation. And I also want to pitch for some Make.com consultation. So now now I have some information about this lead. We have the market industry target audience, their value proposition. And finally, what services are we going to try to pitch. Sure. That we're going to go ahead and just click Create Pitch. Again. That sends a webhook signal to make. It's going to receive that. It's going to put that data together, send it off to ChatGPT and then put whatever it gets back. So here we can now see a text right here. We're going to go ahead and move that down -so we can read it better. -All right. So this is roughly what you will get going back to a bit before about that piece of having isolated pieces of functionality and then having a human check them. This is one of those cases where we go in parts. So step one, obviously fill in the website address that's going to give us some information. We want to still check those just to see that they do make sense. In this case. They definitely do. And then step two we don't do that straight away. We're going to fill in some more information. We do that first check and then we create the pitch. Step three here is really that we're going to take this pitch look at it. And just always consider this as some form of outline. At some point you might get confident enough that it always looks at least to our standards. We'll automatically send that email out. But at the very least here, you don't have to start from scratch. As a as a human. You can just look at this, you can make some slight alterations to it and then you can send it off. So now we actually have a pitch here, just kind of skimming through it. We're going to notice that in this email it's mentioning we understand the unique challenges of the B2B software companies when targeting sea levels. It's actually including some information about us as a company. And it's also going to be mentioning both of those services that we said that we are going to include in the pitch. So further down, we can see by leveraging our monday.com and Make.com consultation services, you can take your business to new heights. Here's how and even gives some, some very clear indications of this is how we can help you in your value proposition. So let's go into Make again and actually look at what this looks -like. -Right. So similarly to before, we have a webhook module, in this case it was triggered when I selected create in this column. Since I want this to be a more manual selection rather than the web address that just go straight away to give us a chance to fill in, in this case, services before starting. After that, it's gonna fetch that entire item. So right here we have that pulse ID, right? ID dynamically mapped again. This is going to retrieve all of those column values that we have about that lead, and then it's off to ChatGPT. This query to ChatGPT is a bit more complex. As you probably noticed, we didn't include that much information from the lead row, but we've actually included this in each and every call instead. So here we have two types of messages. We first have a message with the real system. And we have a message with the role user. What a system role message essentially means is this is for providing some context to ChatGPT rather than ChatGPT, only using what it knows in its collective database. Here, we're going to provide it with some very specific content that we know is going to be relevant for ChatGPT when writing this pitch. So in this case, we've provided some information about who are we, who is Omnitas Consulting. And we've said that it's essentially we work with monday.com, we work with Make. We're an IT and management consultancy firm. This is simply just a bunch of background information that we want ChatGPT to be aware of. Moving on. We're going to look at the user query instead. In this space I've put in act as a sales person. And on this consulting you are writing as the company itself. Keep a professional but still, please. Pleasant, fun loving and friendly tone of voice UK English when you write. Important thing to note there, depending if you want UK English or American English. If you don't specify it tends to be American. Keep the post short. And here we've put in those values that we got from monday.com. So we say write a pitch for potential client working in the. This is the value for the market the value for the industry, the value for the target. So we start off that we want the first part to be just a more general pitch. We're in the middle. We want to include why the following services. So these are the services that I selected that we want to include include why these services are useful. And then we want to end with why the value proposition of the client can be improved by the services of in this case Omnitas, us being the writers. In this case. -We've also included some more specific more, more or less technical or formatting. You can do a lot here, but in our case we want it to be a certain length. So we say no more than 1800 or sorry, 1900 roughly, characters in total. And then at the end, we've again added a very short description of Omnitas, just to really reinforce the tone and the content of that pitch. And then lastly, we do what we did before. We're going to update the pitch content that we got out. And then finally, this is simply just something that is quite nice and visual. As we're going to change that create pitch column that we initially click create. We're going to change that to created to give some user feedback that this went through all the way and we didn't receive any issues. And this is roughly what we end up with. We get a lead workflow that is automated for some information. We use that information to create a pitch along with a manual input stage. -Great. -Any questions? -All right. -So let's have a very -quick walk through again. -We have split this into two separate functions. First part being some simple enrichment. The enrichment is based off the website. After that, we create a customized pitch. The pitch is customized by each client, but the background of the pitch is always the -same. -From here out. A classic use case would be to simply take this pitch, potentially edit it down a bit, changed it around to really make it as you want it even more personal to your needs. And then you can quite simply also within monday.com integrate, for example, your email, whether that be your own, whether that be a group email and send that out to all of your leads. After that, we can use this to actually qualify or disqualify them as well. Question from Erica, could this also be used to potentially scrape other sites, such as Crunchbase for company data? Yes, it could certainly be used for that as way as well. Using ChatGPT or I say scraping tool isn't necessarily the exact same method as a scraping tool would do, but it can oftentimes accomplish more usefulness than a normal scraping tool might do. Scraping tools tend to be more along the lines of. This is set up for a very specific website, for example, Crunchbase, where we'll always looking for the same information and the same space. But I can also be used in this way to do more or less the same thing. Another piece of that is also that generally a lot of web pages that are common to be scraped are going to have some form of scraping protection, whereas the normal fetching like we just did now doesn't necessarily get detected by that scraping protection the same way. So yes, it can certainly be used. And in this use case is more or less what we are using it for some form of scraping functionality. But here it doesn't matter what the website looks like, it's going to provide good information most times, no matter -what the format is. -Great. In that case. Thank you everyone for tuning in. Take care. See you later. Bye.