Taskade: AI Agents for Workflow Automation

The process demonstrates creating AI agents in Taskade for automating research and writing tasks based on blog article titles.

### Leveraging Taskade's AI Agents to Automate Blog Article Research and Writing
* 🤖 Taskade's AI agents can be used as AI employees to **automate tasks and workflows** by assigning them specific roles and knowledge.
* 📝 The first step is to create AI agents with roles such as a "**researcher**" and a "**blog expert**".
* 💡 The "researcher" AI agent is trained by providing it with **background knowledge** in the form of media, documents, and links.
* 🔍 The "researcher" AI agent is then assigned commands to **investigate and analyze topics**.
* 🤖 Once the "researcher" AI agent completes the investigation, the output is used as input for the **"blog expert" AI agent**.
* ✍ The "**blog expert**" AI agent is tasked with generating **SEO-optimized blog posts** based on the **research findings**.
* 🤖 Finally, an automation is set up to trigger the AI agents based on specific events, such as **completing a task with a blog article title**.
* 🚀 The automation then runs the AI agents, and the output, such as the **SEO-optimized blog post**, can be used in further workflows or published directly.

hey everybody. Welcome or welcome back. So, if you're like me and you're still confused about how AI agents work within Taskade and how you can use them to create automations and improve your workflow, then stick around as we go through an end-to-end example in this video of how to create agents and then how to create automations that work off of those agents that we create. By the end of this video, you should have a better understanding and a better grasp of this functionality within Taskade and be able to determine whether or not it's something that you can use in your workflow. So, let's get [Music] started. So, before I get started and dive into the actual details of this end-to-end example, I would like to mention that I do think these AI agents are a very powerful functionality that Taskade provides. However, I do think that the team itself has not done a very good job of showcasing these functionalities by providing real-life scenarios or real-life applications.

do see online there's a lot of people confused about how these really work. Myself included. You know, I played with them a little bit, but I'm still not an expert. So if the task team ever does watch these videos, that is some constructive feedback that I can give, is to provide more examples, more walkthroughs, not just screen recordings. More walkthroughs with somebody actually explaining how to use a functionality in a real-life scenario. But hey, that's just my two cents. I still love the application. And let's get started with this example. So what I'm going to try to do in this n to an example is I want to build an automation or a workflow that, based on a Blog article title, I have an agent that researches the topic around that title, and then another agent that writes an article based on the research and findings of that first agent. So again, the initial input of this workflow would be a Blog title, and the final output that we're looking for is a blog post.

right and with this example, we're going to be using two agents so let's just keep that in mind as we're watching this video. Okay, so again, we're going to start with your our blog post titles. So here I have a project inside my blog folder called blog article ideas, and I have here 10 to 12 blog articles or titles that I want to write. However, I don't want to write this myself. I want to automate and use an AI agent to help me with the writing and the research of these topics. So I'm going to go here on the third tab, the agents Tab, and I'm going to click on create agent. Again, the first agent that I want to create going to be a researcher. And this is a good point point for me to kind of talk real quickly what an agent is. You can think of an agent like a an employee of yourself um, obviously an AI employee that you can assign a particular kind of personality. So in this case, I'm going to create a researcher. And then we can kind

of give it Proms and even give it background knowledge. You know I can start adding links, and I'll show these in a second, but I can start adding links different media that kind of shapes the personality, if you will, of this researcher or of this employee that I am creating so that I can ask it different things. And I know that it it's already kind of trained just as if it was a real person. It's trained based on the knowledge that I provided, and it's going to give me a more detailed or a more, um, how should I call it, like a more precise answer as to what I'm looking for because I have trained the agent beforehand. So really just think of them as AI employees of your workflow or of your company. So here I'm going to use one of the, uh, kind of presets that they have, a researcher. I can give it a different name if I need to and change the Avatar. We won't worry about that right now, and I'm going to click create. Once I click create, kind of the

agent customization popup window appears I see here again the Avatar the name of the agent that I'm creating and here I have instructions so as of right now I'm going to keep it very generic and just use the prompt or the instructions that taskade provides and it just says you are a researcher. Your task is to conduct in-depth research on various topics. Use your analytical skills and access to academic and Industry resources to gather analyze and present data-driven insights. So here in the instructions if you're already using other AI tools like chat GPT or maybe Jasper AI something else to write your articles or to your to the research for you this is where you want to add that template if you already have one. Let's say you're already using another tool to do your research and you have certain prompts or you know certain guidelines that you input into that particular AI tool this is the place where you want to throw those instructions in so that this new AI agent or AI employee of you of yours and follow the same.

guidelines. Here we can also change the tone, and if we need to, we can even choose the output language, which is really cool. We'll just leave it as default. Then I'll go up. I'll go next to the tool sections. This right now just have, uh, web search. I'll go over that in a second, so we don't need to worry about this section right now. Then we come into the knowledge section. This part here is what I think or where I think the power behind an AI agent really lies. So knowledge, again, these are media documents, links, YouTube videos that we can add to this researcher as background knowledge. Basically, we're training the AI agent beforehand with all these data that we're providing so that the output that it gives us in the future, whenever we ask at something, it's more precise and more aligned with what we're looking for. In this case, what I'm going to do this particular researcher, um, all my articles are around work-life balance. So what I'm going to do here, I'm going to add a couple of links to existing

articles around work-life balance that I want the uh to be trained on and use those kind of as a base know. I don't want to copy those word by word, but again I just want to use or have the AI use them as part of its research when it uh researches these topics. I'm going to add a link to two news articles, and I'm also going to add oops about that. I'm going to add a PDF, you know, top 25 jobs for work. Again, this is just another article, but I'm just showcasing here that I'm adding knowledge based on links. I've added two links, and I'm also adding knowledge based on an actual document or a PDF. Here I see the status of these documents or this media. It's ready, meaning the AI has gone through these three sources, and now it's trained or has at least knowledge of these three things that have uh linked Cano in this section. Next, we go into commands, and here we have some commands kind of preloaded for this researcher agent so.

we have the investigate command analyze data report findings. Uh create reports, forecast Trends, etc. If you want to add your own command, you can. But again, I'm going to use the default, the default ones, and I'm going to probably be using the investigate and create. Just investigate 'cause I don't need to create a report again 'cause the, we're going to create a second agent to create the actual output. So for now, for this agent, I'm going to have it just investigate the topic. And for that, I have the investigate command. And this is what the prompt is going to be. Here on this section, I am enabling web search. So this tells the AI researcher to go ahead and search the web if it needs to, when it comes to researching topics. And again, if I need to give more specific instructions around this specific command, then I can insert those here. And I have also toggled here the use knowledge on, as I have already again loaded or linked several sources to this researcher AI agent. So once I have all of these

settings completed. I'm going to click on update, and I see here that my research agent has been created. So now that we have created our researcher AI agent and we gave it some knowledge based on some articles online, and as well as a PDF document, we are ready to create our automation, or at least start to create our automation. I'm going to click here in automation. I'm going to click create Automation, and I'm going to start one from scratch. So again, keeping in mind that we have a project with a list of blog article titles that we want to research and write about, my trigger is going to be when a task is completed. Here I'm going to rename it to blog ideas. So when a task is completed in the blog article ideas, and I want to and what I'm going to choose here is I want to run an agent command. You recall when we were creating the researcher AI agent, we saw that we had different commands such as create report, investigate, analyze, etc., and each of those have their custom prompt telling the AI researcher agent what to

do well. I want to run one of those commands based on the blog article title. I'm going to run an agent command here. I'm going to select the agent. In this case, is the researcher agent. Before I move on, let me just rename this to search. The command that I want to run is an investigate command, and the input for this command is going to be my task content. Again, my task content would be the blog article title from my blogs project. Finally, the last step, and we will change this last step once we create our writer AI agent, but for right now, I do want to see the output of this investigation. So I'm going to create a project. The title is going to be the task content, so my blog article title, and the content is going to be the agent command result. So again, if I go to research, that's my second and a step. I want the result of that investigation to be added as a new project. Select that, save changes, turn the automation on. Just rename my automation to Auto one. Now let's go back to my projects.

so here I have my blog article ideas, and I'm going to select or complete this article achieving work-life balance strategies and solutions. Now open up a second tab here. Going to go to the automations just so that we can see the one we created. I go to runs, and I see here that now I executed a run a couple of minutes ago, and the progress of it is 33%. Go back to my blog folder, and I don't see a new project yet, so the automation has not completed. Okay, now I went back in it. I see now that it's 100%. It says completed. So I do expect a new project once I refresh here to show up. There it is, and the output or the content of this project is the output of the investigation, the research agent did, right? So already we're at a very good start, right? Our researcher is working, and the automation is working based on the article title that we complete. However, we're not done yet, right? This is just the investigation or the initial analysis that one of our

employees did researcher. We want to then create an article based on this research. Let's do that next. So next, let's create our second AI agent. Let me go back here to my blog folder. I'm going to go to the agents tab, and I'm going to click on create agent. This agent I want to be a Blog expert. Since I do want to write an article based on the research and analysis that my previous agent does, I'm going to click blog expert. I'm going to click create here. Very similar settings as the researcher AI agent we just created. Um VI are the blog expert, the specialist in creating SEO-optimized, engaging blog content, etc., etc. This is kind of the general instructions for the writer, employee, or AI agent that I'm about to create. If I want to, I can add more detail as to how many headings I'm expecting, how many words, etc., but as of right now, I'm going to leave it default. Next, the knowledge tab. If I had certain guidelines, keywords, things that I want the agent to keep in mind or be

trained on I can enable knowledge and add those here. In this case, I don't have any, but we already saw this functionality with the other researcher AI agent, so I won't be going through that. And then here we have our commands, and we have here a couple, all around, you know, creating or optimizing blog or written content. But I see here that this first one is the one that I'm going to use the most, which is generate SEO blog post. I go inside of it, I see that it already has a prompt. Let's read that quickly. Create an SEO-optimized blog post on a given topic. Use appropriate templates and incorporate SEO-friendly keywords to enhance visibility. Ensure the content is engaging, well-structured, and provides unique information relevant to the target audience. I am happy with this prompt. I don't think I need to add anything else. I think it's clear enough for the AI agent and should know what to do based on these instructions. And I like that the web search is enabled so that if it needs to

it can go online and search for whatever um it needs. So here the update button is not enabled 'cause I have not modified anything. So I'm just going to cancel out of there, and I see here that our blog expert AI agent has now been created. So now that we have created our second AI agent or writer, the last step would be to go back to our Automation and update it. So if we recall, at the moment, our automation is taking the output of a research AI agent and using that to create the prod project content. However, we do not want this. Now that we have a writer, we want the writer to take the research output of our first AI uh agent and use that output to write an article. So I'm going to go ahead and delete this step. Going I click here on ADD step. I'm going to run an agent command again, edit this, rename it to writer. The action again is we want to run an agent command, but this time we want it. We want to ask our blog expert the command that we want to ask.

our blog expert to perform is a generate generate SEO blog post, and the input won't be the blog article uh post title anymore. The input is going to be the result of the research AI agent, and once I have that, I would need to then finalize the automation by creating a project and outputting the blog post that the writer AI agent generates as the content. If you need to or if you want to take this a step further, you can um connect your WordPress account and configure that as the last step so that whatever output this agent provides will go ahead and post automatically to your blog or to your website. I won't do that at the moment, so I'm just going to create a project title. I still want the title to be the task content, meaning the blog post title, and the content of this new project will be the output but of the writer AI assistant, not the research. If I select that, I'm going to click on Save changes, open up another tab, go back to the blog folder, go back to blog article ideas, let me complete.

another blog title so this is the impact of work life balance on employee productivity. you go to runs. I see here that another run just executed and it's at 25%. so let's give it a minute and I should see here in my blog folder a new article with the title the impact of work like work life balance on employee productivity. orry a new project I should see a new project here and the content would be an SEO optimized article. how the automation is doing at 50% now do wish they had a refresh button here but I don't see one but there it is. now it's at 100%. me go back to to my blog folder and also refresh this and I see now here the impact of work life balance un employee productivity project has been created a few seconds ago and here is the SEO optimized article. kind of short for my liking but again you could probably play around with the prompts and the instructions of the writer. um AI agent and ask it to give you a more or a longer um article.

but hey, if the information is useful, then this is great. So, there you have it. This was an example of how to create AI agents and then use these agents in an automation to output a particular content. In this case, it was all related to content creation around blog articles in particular. But I am interested to see how you guys are using these AI agents and these automations. Obviously, content creation, there's already a lot of tools around article creation using AI, but I still wanted to show an example that's easy to follow and kind of a real-life scenario on how you can create these agents and then use them in automations. But I am very interested to hear how you guys are doing this, maybe if you're even integrating this with other tools like Gmail or Google Calendar, Google Forms, etc. Those kind of bit more complex automations I'm really interested in. So, happy to hear those out in the comments down below. And thank you for watching. I'll see you guys.

Taskade | AI Agents Explained with End-to-End Example

Simplified guides to the latest digital tools in AI, productivity, and online commerce. Learn, apply, and thrive in the digital age.

Progresspresso

Using AI Agents to Automate Workflow in Taskade

Taskade offers powerful AI agents to automate and improve workflow efficiency.

The AI agents can be customized to perform specific tasks, such as researching and writing blog articles based on given titles.

Taskade's AI functionality provides the ability to offload tasks and streamline processes, potentially benefiting users' workflow productivity.

Creating a Custom AI Researcher: A Step-by-Step Guide

Taskade allows for the creation of a customized AI researcher, with the ability to add background knowledge and specific instructions for the AI to follow.

The AI researcher can be tailored to a particular tone and output language, and its power lies in the knowledge section, where media, documents, and links can be added to train the AI.

This allows for more precise and aligned outputs based on the provided data.

Setting Up a Researcher AI Agent for Work-Life Balance

The AI researcher agent is trained on articles and a PDF about work-life balance, using the investigate command to research the topic.

Web search is enabled for additional information, and the agent is equipped with specific knowledge for the task.

The AI researcher agent is created with knowledge from online articles and a PDF document, ready to proceed with investigating work-life balance for the project.

The automation process begins by setting up a trigger for completed tasks in the blog article ideas, initiating the researcher AI agent to perform the investigation.

Setting up AI Agents for Blog Article Research and Writing

An automation process is set up to investigate blog article ideas using a researcher agent.

The researcher agent conducts an analysis based on the article title and provides the results in a new project.

A blog expert agent is then created to generate an article based on the research and analysis conducted by the researcher agent.

Using AI to Generate SEO-Optimized Blog Posts

An AI agent is created to generate SEO-optimized blog posts using appropriate templates and SEO-friendly keywords.

The agent is set to use the output of a research AI agent to write the article and then finalize the automation by creating a project with the writer AI's output.

Connecting a WordPress account can automate the posting of the blog content, but this step is optional.

The Impact of Work-Life Balance on Employee Productivity

Work-life balance significantly affects employee productivity in the workplace.

Automation tools can be used to create SEO-optimized articles on this topic.

There is a growing interest in integrating AI agents and automation with other tools like Gmail, Google Calendar, and Google Forms for more complex tasks.