An AI worker takes GetVape from SEO brief to staging.

A client Workspace. An email brief. Research and website pages the team can inspect. Follow how an agency uses Outloop to organize SEO, generative engine optimization (GEO) and website work—with human direction at each decision.

Start with the result.

Adam opens the returned work, reviews keyword research and inspects a staging page with an answer-first introduction, comparisons and detailed reviews. The output is concrete enough for the team to evaluate and request changes.

Explore all 10 workflow clips

Recorded review of reported research and staging output. Search rankings, AI citations and commercial results are not measured here.

Read the video transcript

All right. So, yeah, we received the email back. So let's see what we have here. the first batch is ready. he did the research for that. you go to the Google keyword tools, and you did actually the research for that. You can see it over here. We have, all the research here. directly from the Keyword Planner. So you can see monthly searchers in United States, the competition, last three months, and then you build on, staging. you build, the new pages. So let's see what is the strategy here. so we have a gap on this category We have no page of them. Okay, so let's see the execution. Yeah, we have a new product. Let's see. So this is the staging environment So this is a image, You have the comparison. I'm pretty happy to have it in this format because this format is the new format of, generative engine optimization. to answer first, then you have imagery over here, and then you have the check prices. Let's see, read the review. we are in the same page. So we create in-depth review for each one of them here. and I just want to see if we create this product in our website. No product at all.

Download English captions

Make the next instruction
easy to act on.

GetVape needed a way to carry search research into website work. In the recording, Adam sets up a dedicated client Workspace, adds existing Skills and configures the worker’s tools and email workflow.

The first research brief asks for a technical check, possible search questions and a proposed page format. It asks the worker to explain its findings before making changes. A later instruction authorizes work on staging for review.

The team then inspects the returned research and pages, discusses improvements and considers which lessons should be retained. The case follows a delivery process; it does not report an SEO performance experiment.

See the steps behind the result.

Choose a short clip. Each includes English captions and a readable transcript. The research-to-staging review is featured above.

  1. Setup

    Create a dedicated client Workspace

    1:05

    Keep the client’s work together.

    Adam creates a GetVape Workspace before assigning tools and instructions. The Workspace gives the team a place to organize this client’s Context, approved access and work records.

    A recorded Workspace setup; each new client still needs its own configuration.

    Read the video transcript

    View all workspaces So this is, Outloop, dashboard. This is the overview s- screen. We wanna create a new dedicated workspace for GetVape. So we will go here, create a workspace, call it Get vape. Get vape, GEO, we can just say go to market Go to market. It will automatically, create the, the workspace ID and, it, it will create it on, on the local computer into the right path, which is on document Outloop project, the new project, right? Obviously, you can change it, but we recommend to use, this path as a default. So we generate, the workspace.

    Download English captions
  2. Setup

    Give the worker reusable Skills

    1:50

    Bring a working method, not another blank chat.

    The session imports existing Skills and scopes them to the client Workspace. These instructions give the worker a starting method for research and website work; the team still decides which methods fit the client.

    Imported instructions are not proof that every described capability has been executed.

    Read the video transcript

    I want to import some skills that I used to work on those skills, on my local computer. So I want to give it to him now, this knowledge to start with. Dedicated, again, to dedicate it to this workspace. I want to click here, choose a file. Import the first one. Done. Sorry. it's nice it detect them me- automatically. So if you import this from, Codex or from, Anthropic and stuff, it detect… auto-detect that, you know, format for you, so you don't need to do nothing specifically. import skills. scope for this workspace. Yeah. Action. Import. Yeah. this is what I want. Done. import Third, tenant creator Import the fourth affiliate links. So each skills has its own knowledge of how to operate this machine automatically, you know, almost You still need a human to interact with this, AI worker on the email, but just tell him like, "Okay, create these twenty pages with animation and optimize them for, SEO and then go to the affiliate website and, generate the right links that we actually get in the, you know, for each sales we're getting a, a commission." SEO optimizer.

    Download English captions
  3. Brief

    Enable work requests by email

    2:08

    Give the team a familiar way to request work.

    Adam configures the worker’s email workflow, chooses an engine and assigns people and permissions. The team can send a brief and continue the discussion in a thread connected to this client’s work.

    The recording shows the configured workflow, not permission for arbitrary senders or accounts.

    Read the video transcript

    I want to go to the AI Worker because from now on, I want to communicate with my agent, from my email. I will choose, Gmail. I will choose the email for the AI worker. So I send to this email, and then I will get response with everything related to this, tools and servers and everything. Choose engine. I want to choose Claude Code as my primary engine for this AI worker. Work owner, I want to choose myself, and I want to add a new one. Add a person. Okay. Sav Affiliate program manager. And I want to choose him as an internal team. I can even add a phone number, but not for now. Available only for this workspace. Add this, person, and then I will go to permission, and I want to edit permission. I want to give him all permission. And I want to give him all the relevant skills, and I want to give him control on the routine. Save permission. There you go. what about myself? I also want to be here. So let's choose myself. add to this workspace Amazing. I want to save the changes. I want to go also to permission, and I want to give to myself, all permission And to save. There you go. and now if I will go back to the AI Worker setup, I want to click enable in practice. So, review and active, email. So you just want to review and active the email, so they allow to send. And there you go. It's, it's running, it's active. So right now I can send an email, and I can get the response back. there you go.

    Download English captions
  4. Brief

    Brief an SEO and GEO research task

    1:56

    Ask for the plan before asking for changes.

    The brief requests a technical check, search research and a proposed page format, with an email explanation and a shareable report. Execution is a later instruction, after the team has reviewed the plan.

    Potential AI-search questions are research hypotheses, not measured query volumes.

    Read the video transcript

    Another email, the GEO and SEO strategy let's speak in a natural, language and just send the brief. Hey, Yoni. So in this task, I would like to, build the generative engine optimization, the GEO, strategy for the brand. So, we would like to understand the technicality. So if we missed anything like a special schema in our website and, if everything is okay on the Google Search Console and GA4, et cetera. just give me a quick report about that. And I would like to build a strategy for GEO. you need to do a research of what people actually are looking for, on ChatGPT, on the LLMs. So you sh-you should try to, predict the, potential searches that will be relevant for our, page. do a investigation and, for, the potentials, queries and searches that the user will search for in order to get the value from our website. go to Reddit, use your, Google keyword search tools. I want to understand, if I remember right, we already designed some, GEO page. So basically, it's answering immediately for, the questions of the users. So I would like you just to show me the format. And, after I will see all the plan, the next email, I will just tell you to, execute it on the website for my preview. So yeah, just give me a report, and it would be nice to see the report also in, PDF, like a design, PDF, so it would be nice to share and to see everything, clearly, but also here in the email, explain to me, in English, short as possible, easy to understand way. Thank you.

    Download English captions
  5. Execute

    Move from conversation to execution

    0:47

    Connect the brief to the tools that can do the work.

    Adam explains how a task thread relates to the worker’s configured website, server and research access. Outloop organizes the client environment; the AI worker carries out the requested work through supported, approved tools.

    The narration describes this setup. It is not a claim of universal integration support.

    Read the video transcript

    just imagine if you will send it to a real employee. all the point here is that this AI worker is not a chat. as you see in the beginning of the video, he actually has access to the server, to the website, to the WordPress, to Google Ads, to social media, to, scraping tools, to, OpenAI model through the API, and so on and so forth. I have something in mind, Just send an email. If it's, a new thread. If it's, existing thread, "Ah, I have this thread about, affiliate strategy. Let's write something." So it's that easy.

    Download English captions
  6. Execute

    Keep website changes on staging

    0:21

    Make review the next step.

    The instruction is explicit: build on staging, let the team review and correct the work, then wait for permission to publish to production.

    A staging instruction is not evidence that production publication happened.

    Read the video transcript

    And yeah, just, create everything on our, staging environment for my review, and then I will review it, correct it, and then I will give you green light for production when it will be ready. Thank you

    Download English captions
  7. Review

    Review the technical findings

    1:07

    Turn an audit into a human decision.

    The session discusses a technical health check and the role of Screaming Frog in finding issues. The person responsible for marketing decides what to act on and directs the next website task.

    No independently measured traffic gain or completed fix is established by this excerpt.

    Read the video transcript

    And we have this insightful, technical health check. So I forgot to mention that also, this AI worker, he has a Screaming Frog. This is a software that's sitting on the computer and doing the technical analysis of the pages, so it can, identify broken links and, pages that compete on the same keywords and, it's really useful. Screaming Frog, he has access to Screaming Frog. it's a software installed in his computer. this worker has a computer like we have a computer. You need to have a dedicated computer, and, it give us all this, in-insightful data. And, it's amazing because you can scan, find a problem, and then, okay, you are the developer also, so fix it That easy But you still need a human to understand marketing and, give him the right, action to do

    Download English captions
  8. Improve

    Save approved learning for next time

    1:09

    Keep the useful lesson after the task ends.

    Adam explains reviewing proposed learning and deciding what belongs in the Workspace’s knowledge or reusable Skills. The team approves what should be retained for future work.

    This is an explanation of the review process, not evidence of automatic cross-client learning.

    Read the video transcript

    I can also tell him, "Okay, look, we did some research over here. Please tell me if there is something that you do not have in your memory or skills and, you need to update." and then it will tell you I have it in the memory, so don't worry." Or it will tell you "You know what? I do not have that part, small part in the memory. approve it to me, and I will save it into my memory, so next time I will, cover it," right? "I will know this." we, build here a important strategy. I'm going to save it in my memory. Do you approve?" If you approve, it will save. So actively, it's designed to offer you, to save it in the memory or to save it as a, capability. So if you walk through repetitive process, so it will tell you "Okay, we did some great job here. You happy? do you approve to save it as a skills?" Okay, that's repetitive capabilities.

    Download English captions
  9. Improve

    Review the work and request improvements

    1:06

    A first result starts the next conversation.

    The closing review identifies ways to change the format, imagery and content before handing the work to the client. The useful result is an editable deliverable with a clear next instruction.

    The video ends with review and follow-up, not a verified production launch.

    Read the video transcript

    So obviously, you can change a lot of here. and you're like, "Oh, okay, I understand this is the best one," and then you will go here to check the price on the website. also we need to send it to the client, and obviously the client will have a lot of, changes but everything is doable, right? We have the execution power. You can send him whatever you want. You want to move the format, you want to add images, to create more product, you can just ask, and you will have this execution power to perform it for you. every marketing strategy and dreams that you have, you can send it to your employee as long he has access. Like a real employee need access, right? for the systems and tools that you ask him to perform on. yeah, guys, I think we done with this tutorial. So thank you for watching and, see you next time.

    Download English captions

One client.
Clear responsibilities.

The agency team
Defines the brief, evaluates research, directs changes and decides what to publish.
The AI worker
Uses the configured tools to research, prepare work and return results for review.
Outloop
Organizes the client Workspace, approved Context, Skills, access and working records.
The client’s systems
Supply the authorized data, website and staging environment needed for the task.

What you can take from this case.

A documented way of working

  • A dedicated client Workspace and scoped Skills.
  • Natural-language email briefs and returned work.
  • Research reviewed alongside a staging page.
  • Human instructions before production changes.

Outcomes still to measure

  • Organic rankings, qualified traffic and conversions.
  • Citations or recommendations in AI answers.
  • Repeatable time savings across other clients.
  • Production publication and its business impact.

The recording presents an answer-first page format as part of the team’s GEO approach. It is not evidence that a particular layout guarantees AI visibility. Google’s guidance says established SEO practices remain relevant to its AI search features; there is no special schema required for inclusion. Read Google’s AI search guidance.

Edited excerpts from one founder-led session. Unrelated inbox content and sensitive fields are obscured. The clips show selected moments, not an uninterrupted audit of every action.

Give your next client brief
a place to become work.

Start with one Workspace, one approved task and a result your team can review. Outloop runs on a company-controlled Mac with your supported AI tools and provider accounts.

Questions about the GetVape workflow