YouTube Video TMjK_m-zlnI ![rw-book-cover](https://img.youtube.com/vi/TMjK_m-zlnI/maxresdefault.jpg) ## Metadata - Author: [[YouTube]] - Full Title: YouTube Video TMjK_m-zlnI - Category: #articles - Document Tags: [[automation]] [[change]] [[implementation]] - Summary: YouTube Video TMjK_m-zlnI - URL: https://www.youtube.com/watch?v=TMjK_m-zlnI Okay. So my name is Akansa. Um I'm currently working as an AI integration marketing manager at RPI. Um and before I begin, I want to clarify one thing that uh whatever I'm going to be speaking about it is from my personal journey uh working in this new role that I transitioned to 6 months ago. Um so yeah so I'm not going to talk about another AI theory that you can get from internet. Uh I will not be talking about what prompt engineering is or what AI is. Um I'll be sharing my real honest journey of integrating AI in the marketing workflows and u this is going to be a developer playbook of bridging the gap. Um so before I talk about what I did, what I what worked for me, what did not work for me, I just wanted to start with talking about why talking about AI is so important right now. Um if you look at the numbers on the screen, you'll see some big numbers. 88% 96% 92%. we see and we get a feeling that we all are in a race. Um, everyone is talking about AI. Um, you if you see the first graph you see that 88% of marketers they say that they are already using AI every day. uh 96% of marketers they say that they have already integrated AI into their marketing strategies and 92% of businesses plan to invest in generated AI in the next 3 years. Um there was a new study that I did not quote here was released by McKenzie on 5th of November that says 63% of U agents are already deployed. Now when you see these numbers it looks like that everyone has figured it out. But when I started with this new role at a uh what was being reported in the news or these social media channels was not something that I was observing and that's because a reality check was happening. Just because you're in a race to adopt does not mean that you are in a race to succeed. Right? Now let's look at the other side of the pictures from the same reports from Mckenzie uh Gartner Forbes. We saw some different numbers. We saw that 53% of consumers distrust AI search results. 62% are saying that the ages that are deployed out of those 63% only 23% have scaled at any level not even the production level at any level and uh uh there was a interesting report came in August this year from MIT that says 95% of genai projects are failing they're failing to meet the promise and goal And if you think about it, there are two pictures. One in one picture it says that we have already succeeded and then the other picture that says that okay uh not many projects or not many ideas that we are working on has been deployed to production level. So what is the gap and what is going wrong? So this leads to me my to my first confession that I'm going to share it with you all that nobody really knows what we are doing with AI. Uh and that's that's an honest confession we should we should not be ashamed of accepting it. Um even though after 3 month 3 years of uh launch uh JAG was launched in November 2022 we are really an expert now using uh prompts or these LLMs but we still feel that we don't understand this new technology. So before we move further we need to accept it that nobody really knows what they are doing with AI. Okay. So before I talk about what I did, what worked, what did not work. I just wanted to talk about my role. Um from my title it says that I am an AI integration marketing manager. So people often ask me like are you a marketer? And my answer to that is no I'm not. I'm actually a tech person. I have worked in Drupal around Drupal for about 9 years and I trans but in at I worked with as a Drupal developer developer work with marketing team so I understand what their pain and what are their struggles and with this new role I'm working to figure out how AI can be used to solve the day-to-day challenges of a marketing person. So enough about this. Let's talk about where does this bridge builder start? What were my first steps? So the first step was not was not a giant implementing a gigantic tool or something extraordinary. It was very simple. Talk to them, listen to them. So I started by interacting with lot of team within the marketing department with a simple aim. Just try to understand what do they really think of AI, what are their concern, what are they excited about and what are they confused about. And I also ran a survey to understand the same questions, which tools are they comfortable with? What are they excited about? What are they confused about? And when I started to work on like when I received the responses from the survey and from this integrations I started implementing integrating AI in the workflow and when I did that I landed to my biggest barriers and I'm going to talk about that and while working on those barriers we have started integrating AI in the marketing team. So the first barrier was that I wanted to talk about is not the technology though it is really it is changing at a very fast pace but the first barrier was not the complexity it was more of the fear of unknown. We found out that the team was not afraid of this technology. They were more confused about this technology. And how do we solve this? we can solve it by talking about it. So we came up with an initiative called marketing AI office hours and in this initiative we fought with this fear fear of unknown. Um in this initiative in this session which we do bi-weekly we kept it very simple. We kept it like we are having a conversation around AI where we would talk about the recent developments, the recent articles of what is happening in AI and in marketing world. For example, if I have to tell you recently charged model 5 and not notebook LM has come up with the deep research and Google labs has come up with an interesting tool called Pom that is going to help small businesses create uh content for their you know for the campaigns and stuff. So we would talk about these articles or recent developments happening in AI world and how does this solve the barrier that we have talked about. By creating an environment where everyone can come together without this this uh this thought that okay someone is going to judge us and we do not know enough about this fast-paced technology. We have created this space where everyone was was feeling comfortable and then we would talk about cover a main topic in these sessions. For example, you talk about what prompt engineering is and you'll tell me okay uh everyone knows what prompt engineering is but what if I tell you there are so many ways proven theories few proven ways where if you apply prompt engineering with those ways you are going to get better output. So we would talk about it. We would talk about craft framework where if you provide context where if you provide a role to your uh in your prompt wellm uh an action item what format you're expecting and a target audience then then you are going to get better output. If you have to brainstorm on new campaign ideas you cannot just provide a vague prompt. If you use treebased technique then you are going to get better output. So we would talk about such things in these sessions and in the language a marketer is familiar with. So once we did that we we were on to our next barrier as in what more we can do. So what we did when we started bringing AI when we were thinking of what more we could do we landed onto a second barrier which was about we don't have enough time and you all can back me on this learning a new full-fledged tool requires enough time and we all are busy marketers we all are busy person we don't have enough time to learn a full-fledged tool if you if I have to say that do you know all the features of Google Gemini Nobody knows even I don't know if you know all the features of charge what charge has to offer. We know we can prompt we know we can use deep research with but there are so many more things that you can use. So the second barrier was that we do not have enough time and if someone has to bring AI into their integration into their workflows into their life or into a team you cannot start by integrating a full-fledged tool. You have to start slow and that's exactly what we did. So what we did, we started slow. We started slow such that we are not disrupting the existing workflow of a marketer. We are we we identified the use cases where the team is already familiar with the team is already confident with and we identified the use those use cases and then we worked on them. For example, this teeny tiny small use case which is often overlooked writing all text for images and I read it somewhere that um from the gardener that says that you lose 17% of your traffic if you if your site does not implement accessibility and writing good all text is is an important part part of maintaining accessibility on site and this is something which is often overlooked can be can be you know implemented better with AI and this is a use case that we have started with. If you see we have simply added a button we call generate with AI in the Drupal media library. Nothing fancy nothing big but it solved us provided us great uh starting point. This helped us build the massive trust and prove that we were listening to the team. We are not starting with something gigantic. Next once we did that this team and small integration, we were like okay what more we can do and if you know this has also talked about how easy it is to build content with Drupal now thanks to the Drupal AI initiative and that's exactly what we did. Then we integrated AI assistant into our um content creation workflow. Nothing fancy again. It is really easy to integrate AI now. Just install a module and then you'll have this assistant where you can integrate it with 10 different LMS. With this um marketers were able to create content. They were able to change tone. they were able to um uh translate content. They were able to fix the styling and structure without moving tabs, without copying and pasting uh content from different LMS to the Drupal site. Now again what it helped us achieve? It helped us achieve first we were we were moving ahead in our integration AI integration journey and then it is helping us helping the team getting comfortable with AI. Once we did that we were like what more we could do. So we brought AI into our translation workflow. Again re talked about today only how translation is important that in Belgium you have to publish content in three different languages. So translation is an important part and if you bring AI into translation of course it'll automate it will do good it'll help you achieve uh things in fast fast way but one more thing it'll do what if I tell you integrating translation into your workflow is going to save you money as well so let's say if you have a website with 100,000 words and if you have to translate content of that website. Um, of course, you can implement this uh interesting tools like TMGMT and pair it up with external vendors like Google Translate Smartly. But if you are going to translate it with those vendors, you are you are going to spend $6,000 for translating 100 uh,000 words. But if you have to translate it with AI, you'll be able to do it with just $2.70. 99.95% of saving that we have done here and and once we did that we identified one more barrier. The barrier is low quality. We can translate content with uh LLM and we can generate content with LLM but we cannot 100% rely on it. LLMs are bound to hallucinate a lot. A lot of research is going on this hallucination but still they are hallucin hall hallucinating a lot and in the survey also we found out that 23% fear that AI would create bland content that sounds robotic unnatural and does not meet your brand. So how do we fix it? How and if we are just going to implement AI assistant thanks to AIA initiative and have published the content on the website we cannot rely 100% on it. It is not going to reflect your brand voice. So we solved it by bringing more not just Google AI assistant. We we went ahead with an hybrid approach. We introduced Jasper. Jasper is a content creative tool where you can where if you use it. I'm not going to talk about more Jasper but I'm going to talk how it helped us. So the problem was that we cannot rely on the content generated by LM it was not meeting brand voice. But with Jasper bringing this hybrid approach in the translation workflow, we were able to decide um if we have to create content that is going to match the brand voice, we we can use Jasper. If we have to create content that is going to meet the targeted audience, we can use Jasper. And remember whatever we were doing, we were making sure that we have human in the loop. We were applying safeguards. So once we do did that we were like what next we can do. We have worked on the trust we have worked on the fear we have worked on integrating AI in our workflows and in the translation workflows from small to large workflows but what next we can do. So once we like okay we have done enough next we can probably work on a full-fledged AI system and we decided to do that only but we were thinking about okay what if we fall in the 95% criteria you remember in the second slide we talked about how MIT has published this study where it says 95% projects they fail to meet the end goals. So the barrier number four is that you can build stuff but it is going to fail if you're not following the right steps and the right steps are first work on the foundation first work on creating the environment where everyone is equally contributing towards AI not just one toward or a dev team and once you do that start small apply safeguard and we are doing that only and we are working on an interesting we saw an interesting opportunity where we can simply um integrate AI and build an AI system not just simple integration and we are working on this great initiative called autodiscocovery uh BDR AI system. This is a system that is going to help our BDRs prepare report before their final meeting. prepare reports on the accounts or uh target company get do its own research cannot rely 100% on AI as I've mentioned earlier uh get information from 10 different uh platforms for example people's lab for examples data for SEO crawl their websites get information with platforms like file crawl and integrate it in a system called autodiscocovery media we are still working on it so of course I cannot not show you how it looks like and what sort of report is it going to generate but I built a demo. If you remember Dre has talked about active pieces he talked about this low code no code solution that if that that is going to help you automate your workflows with with uh with AI in minutes and I did exactly same. So what what I did I used this no code low code solution called active basis. Why there is one more actually two more famous no code low code solutions called zap and n but I chose active species because it is open source it matches the Drupal energy and I built this I tried building okay what can I demo in this project sorry in this pres presentation. So I like okay there is this common problem that all the marketers face. You can go and create content with the help of AI assistant. Um but once your content is published then you have to create content for different social media channels LinkedIn, Twitter, Mailchimp and for that also you can go of course use these LLMs but there also multiple iterations copy content and then publish them. This is a manual step. This is a manual work and you can automate it. You can automate it with the help of these low code no code solutions. And why am I emphasizing on low code no code work? Because it is very easy now to build these automated workflows. No technical deep expertise is required. I built this workflow within 15 minutes. So let me show you what it is and then I'll talk in the tool we all know we go and create a new starts in the tool we all know and love Drupal. We go and create a new content. This is going to be the title. I'll use AI assistant to help me create the content of this article. I'll paste the structure of the article. I'll say use this structure to create a SEO and AU friendly article. Now AI will do the job. Create the article for me. It may take some time. It is generating content and it is done. We'll go save changes to the editor. I think I'm good with the article content. I'll go and publish it. So, it is published. In a normal workflow, this is where the pain would start where we have to open 10 new tabs for creating the content for different media channels, LinkedIn, um um Twitter, Mailchimp. But now with the help of this workflow, we should see a notification in our Slack channel called articles approvals. I'll refresh it. I should see that it says her hey team a new post is ready for review where I can see the draft for LinkedIn post and Twitter post. Now it is up it is on me whether I need to approve it or disapprove it. If I disapprove it, it'll stop the workflow. If I approve it, it should go and publish the article to the respective social media channels. Let let's try with approve. Now I'll close it. I should see a published post on my LinkedIn. Okay. Yeah, it took some time and it says, hey, approved a live experiment connecting Drupal AI active species has been forced into LinkedIn. If I refresh it, I see the post on my LinkedIn. Um, cool. So, uh, I was thinking to demo it live, but unfortunately something is not working with LinkedIn API on active pieces. I prepared the content that okay, I'll demo it live, but it is not working. Uh, it's okay. But I just wanted to show you how this simple manual uh this simple workflow can be done implemented in minutes. So this is how it looks. Right now it is working with LinkedIn but you can integrate different social media channels within the same workflow. It will work with Drupal. It is shipping removing unnecessary tags quering the chat GPU to prepare the content. Again the content is is coming in JSON form. Uh now change it to the form that is appropriate to send to Slack. Once you do that send approval request to uh Slack. This will trigger approval request on on Slack where human is still in the loop. Make sure human is always in the loop. Once the human is is in the loop, we have route where it splits uh based on what the approval stage is. If it is approved, it will publish the content to LinkedIn and then notify everyone over Slack. I built this in 15 minutes. So, building stuff is very easy. Now, it is not that uh you have you have to know lot about technical stuff to build workflows. Now, once you do that, let me share again. Okay. Yeah. So once you do that, you'll be able to build something that is not going to land in that 95% category. You are going to build something that is going to scale. Why? Because we have worked on the foundation. We have worked on on on something where the team is team is coming together and building it together with you. >> Now we are on the last last slide. uh I just wanted to wrap everything what I've talked about and built in this last 6 months in in a framework where if you have to go back and talk about it that how you should bring AI in your team you can do it with four simple steps create an environment where you're talking about AI where everyone is equally involved in AI and once you do that once the team is feeling comfortable confident You can start small. Don't bring gigantic tools and expect team to be fluent in it. Start small. Identify use cases where you can bring AI, integrate AI. And once you do that, apply safeguards. You can you can u keep human in the loop. But and if you're confident, if you have a use case where you are feeling okay, I'm not going to simply integrate AI. I'm going to build my own LLMs. You can go ahead with something uh like building or deploying your own LLMs on services like Amazon Bedrock but make sure whatever you are doing you have safeguards enabled and once you do that your foundation is strong enough you can work on something big measure learn and then scale okay I'm on the last last slide um that's all from my end uh there's is this QR code where you can scan it and it will take you to my LinkedIn profile. Uh whatever I talk about I like uh sharing it on my tech journal. Um and I like sharing it on my YouTube channel also because I feel the best way to learn something is when you share it with others and that's what Drupal community is for. Uh and that's why we all are here. So I share it on my YouTube channel also. Uh and the last link is to my Drupal profile where I share and list all the interesting work I'm doing in Drupal. So yeah, up for questions if you have. Perfect. My job is easy now., YouTube Video TMjK_m-zlnI, YouTube, 01kabyckk654g6pev9mka9epp3.