I Taught My AI Agent to Write in My Exact Voice. Here's the Playbook.
Prefer to watch? Watch on YouTube · Subscribe for new builds · Read the transcript
You know that feeling when you ask ChatGPT to write something and it comes back sounding like a LinkedIn recruiter wrote it? "In today's rapidly evolving landscape, we'll explore the fascinating world of..."
Nobody talks like that. I definitely don't talk like that.
So I fixed it. And it took me about 10 minutes.
The Problem With AI Writing
I run an AI agent called The RZA on my Mac Mini. Yeah, like Wu-Tang. He runs every morning at 6 AM and writes four video scripts for my YouTube channel. Fully automated.
When I first set this up, the scripts were fine. Good topics. Good structure. But they didn't sound like me. They sounded like AI. And that's a problem because I read these scripts on camera. If it doesn't sound natural coming out of my mouth, people can tell.
You've probably experienced this yourself. You use AI to draft a blog post or an email and then you spend 30 minutes rewriting it so it doesn't sound robotic. At that point, why even use the AI?
The Fix: Build a Voice Profile
Here's what I did. I took three of my best YouTube videos. Videos where I was just talking naturally. Being myself. No script, just vibes.
Then I transcribed them using Whisper. That's OpenAI's speech-to-text model. I ran it locally on my Mac Mini. Downloaded the videos with yt-dlp, extracted the audio, ran Whisper. Maybe 10 minutes total for all three.
Then I went through those transcripts and pulled out everything that makes me sound like me.
My signature phrases. Things like "y'all" and "which is pretty cool" and "holy moly" and "it's a skill issue." The way I start sentences with "So" all the time. How I say "you know" as a connector between thoughts.
My sentence structure. Short sentences. Fragments. I don't do long complex sentences. That's not how I talk.
My energy. I'm genuinely excited about what I'm learning but I'm not performing excitement. I'm self-deprecating about my coding skills because I'm still learning. I say things like "I had no idea what I was doing" because that's true.
How I explain technical stuff. I always break it down simple. Like I'm explaining it to a friend. I use analogies from real life. And I admit when I didn't know something.
I put all of this into a document called beau-voice-profile.md. That's it. That's the whole trick.
Before and After
Before the voice profile, here's what a script opening looked like:
> "In today's video, we'll be exploring the fascinating world of AI agents and how they can transform your workflow."
I would never say that. Ever.
After the voice profile:
> "In this video, I'm going to show you something that changed everything about how I use AI agents. And I'm not going to lie, when I first figured this out, I was pulling my hair out trying to get it to work."
You hear the difference? One sounds like a robot. The other sounds like me talking to you.
The Voice Profile Structure
You can do this for any voice. Doesn't have to be yours. Here's the structure:
1. Opening Pattern How does this person start their content? I always start with "In this video, I'm going to be going over" or "If you're trying to do this thing, this video is for you."
2. Signature Phrases The specific words and phrases they use over and over. For me it's "y'all", "which is pretty cool", "So," to start sentences, "can I get an amen." The things that make you sound like you.
3. Sentence Structure Short or long? Fragments or complete sentences? Do they ask questions? Do they use rhetorical setups?
4. Energy and Tone Am I a guru lecturing from a mountaintop? No. I'm a builder sharing what I'm learning. Peer to peer. That energy matters.
5. How They Explain Things I always simplify. I use analogies. I admit when I didn't know something. Very different from someone who explains things like a textbook.
6. What They NEVER Do I never use corporate buzzwords. I never talk down to viewers. I never do fake energy. Knowing what to avoid is just as important as knowing what to include.
The Banned Words List
On top of the voice profile, I have a humanize guide. It's a list of banned words and patterns. Things like:
- "delve", "leverage", "landscape", "comprehensive", "robust"
- Perfect parallel structure (real people don't talk in perfect triads)
- Staccato LinkedIn-style writing
- Hashtag stuffing
If a word screams "AI wrote this," it's banned. Plain and simple.
How It Works in Practice
Every morning at 6 AM, The RZA wakes up. Reads the voice profile first. Then reads the daily intel from another agent called Ghostface who gathers trending topics. Then checks the database for what videos we've already done so there's no duplicates. And then writes four scripts that actually sound like me.
I wake up, I've got four scripts waiting. They sound right. I copy paste into Google Docs, read them on camera, and twenty minutes per video. Done.
The whole pipeline is automated. The only manual step is me sitting in front of the camera and reading.
Try This Yourself
Here's what I'd recommend if you want to do this:
1. Record yourself talking for 10 minutes about something you know well 2. Transcribe it (Whisper is free and runs locally) 3. Pull out your patterns. The phrases you repeat. The way you structure sentences. How you transition between topics 4. Put it all in a document 5. Tell your AI to read that document before creating anything
That's it. Not complicated. But almost nobody does it.
The difference between AI content that sounds like AI and AI content that sounds like YOU is literally a 10-minute exercise. No fancy tools. No expensive courses. Just transcribe yourself and teach the AI what you sound like.
If you want to learn how to actually set up AI agents that work for your business, not theory but real setups like this, come check out Shipping Skool. Over 50 members building together every day. We cover everything from agent pipelines to voice profiles to automating your entire content workflow.
Be blessed, y'all.
Transcript
Read the full transcript
0:00Over the past 4 days, OpenClaw has generated over 80,000 views and over 2,000 subscribers on this YouTube channel, and I'm here to break down the entire content pipeline and what I built and how I'm automating everything. So, every morning at 7:00 a.m., I wake up to a completely fresh set of content, podcast outreach, newsletters, blogs, four YouTube scripts with thumbnails, five social media posts, a full blog post, and all created by AI agents that kick off on cron jobs every single day. I run it all on the Mac Mini behind me. So, I have no monthly subscriptions other than the Claude Max plan, so I have no cloud costs. I'm not waiting for API calls, just a team of AI agents that work while I'm asleep. So, basically, they're turning trends into data and
0:45then turning that into ready-to-publish content. I also do need to review everything that is being done because sometimes things can go off the rails, and I do need to check the work that's being verified. But, this is saving me at least 10 hours every single week. So, I'm going to tell you exactly how this works, the real dashboard, the actual agents, and then the numbers behind them running the automated content business on local hardware. The whole system is called Mission Control, and if you're new to OpenClaw, this is basically just a dashboard where you can see everything on what's happening visually. It runs off Next.js in architecture, and I built that, and it all ties together with the multiple agents handling different jobs all within OpenClaw. Each agent has a
1:31specific job, and they all work together like a content production team. So, here's how it works specifically is at 5:00 a.m. every morning, an agent called the Ghostface wakes up and starts gathering intelligence. By the way, all the agents are named after Wu-Tang characters, so I just love it. I love the Wu-Tang, and it just felt natural. So, it scrapes Twoblab for trending YouTube topics. It monitors X for viral posts, it checks Reddit for discussions, and then pulls the data from Hacker News. And then at 6:00 a.m., we got the Rizza takes all of that intel, and then it generates four complete video transcripts, outlines with titles, thumbnails, all for me to pick and choose. So, the are these aren't just
2:16like topic suggestions, these are actually relevant news articles based on past content, and everything's ready to go. So, everything's written in my voice because we pulled transcripts from other YouTube videos that I have on another channel. I fed it all my emails, all my tweets, so it sounds just like how I would sound. So, 6:30 a.m., You Got generates the custom thumbnails for each script, and it uses Nano Banana 2 through the Kai.ai API, and then You Got also uses TubeLab to look for outlier scores. So, it knows my face, it understands the topics, and it creates the thumbnails that match each video concept. At 7:00 a.m., a separate blog agent writes the full article based on
3:02the video topics, and Inspect Deck writes the five social media posts to promote the content. All of this happens automatically in the background. So, I basically wake up and I get to check Mission Control, and I have a full day's worth of content waiting for me. I also have an agent, Method Man, he takes the videos and turns them into clips. Don't need to pay for a clipping service if you can build it out yourself these days. So, why did I build this? Because content creation was killing my productivity as a builder. I was spending 3 to 4 hours every day just on YouTube writing scripts, creating thumbnails, planning the content, and writing the social media posts. That's about 15 to 20 hours every single week. So, here's the thing, good content
3:49drives business. YouTube brings in subscribers, subscribers join the community, and now the community is at 56 members in just 5 days. So, YouTube got monetized just yesterday at 3,200 subscribers. Thank you guys for watching. If you like AI news and AI updates following my journey, make sure you subscribe because we got four or five videos coming out every single day. And yes, this is me. I'm not a HeyGen Avatar. So, the content machine just works and I built it to to buy back my time. That is my main purpose for using OpenClaw is time is so valuable to me and I was looking at all the things that I was doing every week and I was like, "How can I switch out all of the redundant tasks that I do every week and
4:34have AI agents do them for me?" And if you're getting going with OpenClaw and you want to build out something similar, I would start with one automated task first and then build out your agent workflow from there. So, I looked at what I needed to automate and I built it out. It did take me over 30 days to get this thing actually working. So, if you don't have patience, you know, you're going to learn it getting this set up. So, that's where this exact system came from. I wanted to keep the content flowing without consuming my entire day. So, I'm going to talk a little bit more about Mission Control. This is basically the brain of the operation built using Next.js and it just runs all locally for the Mac mini that I have behind me. So, I could see what I have going. I have several tabs breaking down all different
5:20parts of the operation like Trend Radar. I have a videos tab, a shorts tab, calendar, and overview. The Trend Radar shows all the intelligence Ghostface has gathered like the trending topics on YouTube, the viral posts on X, hot discussions in subreddits, and the data feeds the content generation process. The videos tab shows today's generated thumbnails, scripts, different hooks, different options. I can look at this and choose which one I like and go with it. So, it's just ready to go. I can read through them and edit if I needed and basically just start filming. So, each script comes with a custom thumbnail that's generated by another agent called U God. These aren't just generic templates. The agent actually uses the Tube Lab skill and the Tube Lab
6:07API to actually pull outlier scores. So, it can understand the topic and then recreate the thumbnail that match the content. So, the blog posts, it's generated every morning. You get a full article that's SEO optimized, ready to publish. I have the site map on the Google search console and I have it being crawled every few days. So, all of these blogs can get fed into the SEO pipeline so we can start ranking on these topics. So, here's the five social media posts that Inspector Deck creates. They're not perfect and I do need to go in to alter every single one. But, like I said, it gets me about 80 to 90% of the way there because it just needs a little bit more of a voice and a little bit of more, you know, perspective. So, it it does good.
6:53It's just not there yet. So, the nice thing is it's all created while I'm, you know, enjoying my morning cup of coffee. I don't have to have any human intervention. So, now let me talk to you about like how the agents actually work. This is an open claw. This is open claw running on my Mac mini and you can see the agent squad is basically about eight agents doing all different tasks. Each agent has a specific role and a set of tools. So, Ghostface is the intelligence gatherer and that's what kicks off every morning. It also, you know, doesn't just collect like random content. It's looking for specific patterns, topics that are like getting engagement, stories that have momentum, discussions that indicate interest from my target
7:38audience. The output from Ghostface feeds directly into the Rizza. So, Rizza can then take that and use it as like a base for script generation and agent takes all the intelligence from Ghostface and then creates a complete video outline. So, it's pretty cool on what you can get going nowadays and these are the things that I would spend hours a day on building out just for one YouTube video. Is it perfect? No, but it does get me a lot further than not having it, you know? So, the Rizza does know my voice profile and understands like how I talk, what examples I use, how I structure the content. It also references my actual experience with Open Claw. So, when I'm building with Open Claw behind me, it can take the daily logs and actually weave it into
8:23the content. So, the Mission Control Shipping School, it doesn't just fabricate stories or claim I've used these tools I haven't tested. It's actually has a truthful concept of what we're doing, especially when it comes to community members in the Shipping School and how we've helped them in gathering testimonies. That's the thing is this channel's called Building in Public. We're going to leave nothing left behind. We're going to show you how I'm building this, what I'm building it and why. And that's why I want to be 100% transparent with the things that we're doing. So, the cool part is like here's an example like Ghostface yesterday found a trending topic about Apple's new AI features and then the Rizza took whatever it learned and it wrote a complete script analyzing the
9:08announcement comparing it to local AI solutions like Open Claw and then tying it back into what builders can learn from Apple's approach. The script was ready to read on camera. I just had to review it and start filming. You got to handles the thumbnail creation and it actually does a really good job than what we were doing before. So, it basically downloads the thumbnail from the Outliers and then recreates it and generates a prompt to feed back into Nano Banana. It has my face as the reference photo, so it has that training data, and it understands the visual style of my channel, and then it creates thumbnails that match each topic. And the thumbnails, you know, they aren't just random AI art. It actually analyzes the script content, and then it creates visual that will support the message. If
9:53the video is about automation, it might show me with a confident expression and tech elements in the background. If it's about breaking news, it might show surprise, excitement. And Inspect the Deck, you know, does great. I think we need a little bit more refinement when it comes to the X posts. These posts, it does analyze proven engagement patterns, so it will look back through my X account using the Grok X API and see what did well, what hooks sucked, and it's kind of like this self-refinement process that gets better and better as we use it. So, Inspect the Deck, you know, will will write follow-up posts and reply templates, though I don't have it just fire off automatically. In fact, all of the things that are written and pre-done, I have to auto-approve, which is just keeps it much safer, and I can always
10:38review things, so it's really nice. I have a little bit more control. The cool part is just basically these agents all work together, and it all happens in Mission Control. The dashboard coordinates everything, and it stores all the content in Neon database, and gives me a single interface to review and approve everything. So, I'm going to talk about the real numbers behind the system. So, since implementing this automation, the content output has increased 3x. I'm publishing daily instead of two to three times a week. The quality hasn't decreased because the agents are trained on the high-performing content patterns and use my personal voice. YouTube subscriber growth has been insane. Four days ago, I had 1,000 subscribers. Today, we have a little over 3,300. We just got monetized, thanks to you, and that's an
11:23additional revenue source on top of Shipping School. Speaking of the community, we are now at 55 members, generating over $1,400 of monthly recurring revenue. The automated content funnel is actually working. People discover the channel, they see the value that we provide, and then they want to join a community because if you're a builder like me, it's good to be around another group set of people that can share ideas, help you build your project. We run four bootcamps a week. I also provide one-on-one help to help you get OpenClaw installed. Maybe you want to build out a content generation pipeline just like mine. Well, we will help you with that. Maybe you want to learn how to use CloudCode to spin up your first app idea. We offer a starter pack for that. So, that's the cool part is we're providing as much value as we can at scale and at volume, and I'm just
12:09one person. So, my time investment into content creation has dropped from 20 hours a week to maybe 3 to 4 hours. I spend time reviewing the scripts, filming, and engaging with the comments. The heavy lifting is automated, and I get to spend more time with my community actually working with them one-on-one or in a group-like setting. And the cool thing is is basically the cost the cost to run this whole system is just $200 a month. You know, I paid for the Mac Mini, that's fine. At the time, they were $500. And OpenClaw runs locally, so I have a bit more control. The only reoccurring cost is like the API calls that we make with OpenRouter, which is like $10 a month, and as well as using the kai.ai API, that's a mouthful, but that costs about $50 a month because I
12:55am generating content based around my three SaaS apps, EasyFlip, Magic Hand, and SnapTastic. They all do different things, but I'm able to pull viral content off of TikTok and Instagram, recreate that content using the Sora 2 API within Kai, and then it's able to overlay any information that needed, and then post that content on my social handles, driving people to the website, and getting users. That I will go into on another day, but I just built that the last few days, and that is also saving me a ton of time. So, compare that to what I was paying before, and not only just my time, but the Claude API costs, and you know, thumbnail design tools, and just doing it myself
13:40added up. I was spending hundreds per month with the content creation tools. I used to have Canva. Um there is other consulting agencies that I used to have, you know, monitor my channels, and that was $100 a month. I used to have video editors, which was a couple hundred dollars a month. I used to pay for Opus Clips, which was $50 a month, and I built all that out on myself, you know, free. So, that's the cool part is you can actually save money cutting costs not having to pay for several different subscriptions. I guess the key insight to all of this, if you're wondering what I'm even talking about, is just about saving time and money, and more importantly building a system that scales without me burning out. I love creating content. I love providing value. Most of all, I love helping people like yourself do the same
14:26exact thing, buying back your time, doing the things that you're passionate about, and that's what I'm making these videos for, are people like yourself. So, before automation and increasing the content output meant working more hours, which I didn't have, now I can scale the content without scaling my effort. The agents work 24/7. They don't get tired. They don't need breaks. They don't have creative blocks. It's a recursive loop. These agents get better and better and better as we go on. The system has fundamentally changed how I think about content and business operations. Instead of being the bottleneck, which I was, I became more of like an orchestrator and a manager. I set the strategy review and the output, and then I get to focus on
15:12the high-value activities. The content quality stays consistent because the agents follow the proven patterns and the frameworks. For example, we have an agent that looks at my YouTube channel every week. It says, "Okay, these thumbnails performed. These videos got the most view. Let's recreate those topics. Let's talk more about Open Claw, Claude Code, and let's maybe skip out some of the AI news that you don't really want to know about." So, those are the things that we have going on in the background. Like I said, a loop that just feeds back and things get better and better as we go by. So, every script can follow the same kind of engagement principles that have been working, and then ditch the stuff that sucks. Now, you might be thinking this sounds too good to be true. But, on
15:57shippingschool.com, if you go on there, there is a use cases tab with over 60 documented use cases on how people are making money with Open Claw. You could click the link down below, check out the use cases, and read it for yourself with verified sources. So, I'm going to address some of the common concerns. It's like, "Doesn't the content feel robotic or generic?" This is why I auto approve everything, and I have to go through, you know, reading some of the stuff because it's not perfect. Nothing is going to be like, "That's amazing. That's a perfect video idea." I think you're going to need to review things for at least the foreseeable future to be able to add in your own two cents on, you know, maybe what you think about it, you know? And that's a common concern, you know, people don't want that. Because the AI
16:42agents are trained specifically on my voice and style, they reference my actual projects, experiences, and opinions. They don't fabricate the stories or claim expertise I don't have. I'm not claiming to be some type of guru or you're not going to sell you a $40,000 course like Tai Lopez. I'm just going to show you everything that happens every single day on the mission control, which is right behind me, where you see me pointing. Um, and it's basically going to just have these logs. And if I made a mistake, or a cron job broke, or I ended up switching out a for my Oprah story agent, that broke the whole system. You're going to know about it. So, that's the cool part is that it's like a daily log of what I'm doing every day and how I'm doing it. And also
17:28with the community members as well in our calls, that transcript gets picked up and also weaves it into the content. So, the content can feel more authentic because it is authentic. It's the things that I'm actually doing every single day. It's just automated. So, instead of manually creating this and doing all this, I can just have all these happen automatically in the background. And you might be wondering, "Can agents handle complex topics or breaking news?" Like Ghostface monitors the trends in real time every morning and then the Rizza can pick up on that information and generate content around it. So, the due developments that come out, like Open Claw just dropped a new update last night at 9:00, I have it trigger any Open Claw update that comes out within
18:14minutes. I have a script. I have it updated as well. Anytime an update comes out, I have it update automatically. And then I have thumbnails done. So, anytime Open Claw releases an update within minutes, I can break the news, tell you about the news features, and how I'm using it. So, that's the cool part is we're on top of breaking news. And things in the AI space happen super fast. So, what happens when something needs human judgment? And this is a common thing, too, is that's where the approval process comes in. Like everything goes by me. I read everything that can be published and I need to know what's being published. Nothing goes through automatically. So, in mission control where I can review and edit or reject it, I can get much more control out of it. So, the automation handles
19:01the heavy lifting, but I maintain editorial oversight. Here's the what this means if you're trying to create content or you're just trying to grow your business and maybe get some more eyeballs on the product you just launched. Content marketing is one of the best ways to build an audience and drive business results. These modes are vanishing and being able to content-led founders are going to be more and more in abundance in this coming year because people want to do business with people who they know, like, and trust. And if people don't see the face to what you're building, they're not even going to know what you're doing and who you are. Why should they trust your product? So, content-led marketing with founders in the loop is going to be huge, especially this year. And it may take time for you
19:46to realize that. I'm not saying you need to get in front of the camera and talk about the products you're doing because now with, you know, Sora 2 or VEO 3.1, you can actually copy viral content and put it your own spin on it. It's not um, you know, it's not it's it's one of the most time-intensive aspects to running a business that people just were like, "I don't have time to do that." But now more than ever, there's no excuse. The system proves that you can have both. You can maintain consistent content presence without sacrificing your building time. Cuz I love building, I love creating things. That's just me, it might not be for you. But you can scale your content without scaling your effort and feeling like you're burning out. I think the key is thinking systematically and like instead of manually creating each piece of content or thinking about what direction you want to take it, you
20:31design the systems that create the content automatic. You identify the patterns that work and then discard the junk that doesn't. You get to train your AI to just to follow these patterns, get better, and produce the results. So, you build the dashboards to coordinate everything and then you maintain quality control through the review processes. It really isn't just about content, either, but I just wanted to share this with you because this is what's been changing over this last week. It's a, you know, I would say the same principles apply to customer support or like any kind of lead generation system or if you're doing product research, market analysis. So, anywhere you have repetitive knowledge work, you can build systems to automate it. The future really belongs to the builders who know how to orchestrate AI systems, not just use
21:18individual AI tools. So, if you want to learn to build systems like this, this is exactly what we do inside of our Shipping School community. I don't just teach you about the AI tools, I work with you to build the systems that solve your real business pain points. We do four live calls per week and I walk you through the setups, like getting your mission control going, how many agents you should have running, who should be the orchestrator agent and what models you should be using. So, you get basically hands-on help building your own automation pipelines. You get to learn tools like Open Claw or Claude Code and other tools to create systems that actually work for your specific businesses. The goal isn't just to automate the content, it's to automate everything that doesn't require your unique human expertise. So, content
22:05creation or lead qualification or customer research, competitive analysis, things like that. You're going to get to be able to free up your time to focus on building products and serving customers. So, inside the community, I'll help you identify the automation opportunities in your business and build systems to capture them. We start with your biggest time sink and then we work backwards from there. Today, in just a few hours, we have our Ship and Roast Bootcamp where people can bring their projects and we get to roast them and see how we can make their product even better. The members I get the most value are the ones who come in with a clear problem they want to solve. So, maybe you're spending too much time on social media, maybe your customer support is overwhelming you, maybe you just want to start a content strategy, but you do not have the bandwidth. I've been there,
22:50trust me, I burned out on social media. We've been making videos for 6 years on other channels. There's a huge burnout cycle. But, basically, we take that problem of burnout and then we get to build a system that solves it together, Step-by-step, call-by-call, until you have an automation working, a you know, that's the goal. So, the future is builders and creators who understand these systems and leverage these AI tools to get that outcome they're looking for. And these systems compound over time, especially if you have a recursive feedback loop that improves itself over time. That can be very powerful. So, Mission Control basically just started as a simple, you know, content generator, but now it's a complete content business running on autopilot. Most people pay thousands of dollars a month just to generate
23:35something like this. Not anymore. That's the power of thinking systematically about automation. So, if you want to start building this, I'll put a link down below, you could check it out. Over 50 members in just 4 days. So thankful, and I love helping people most of all. I would say that is my biggest passion. So, if you're on the outside looking in, it's not too late to get started, but I would encourage you to be around like-minded people that can hold you accountable to actually build the thing you're looking to build. So, I would say the creative decisions and the strategic thinking and the relationship building, this is the biggest thing. You could free yourself to do other things like high-value activities. You know, I have other businesses running in the background that don't require as much time as any as it used to because the systems are built and put in place. So,
24:21basically let AI handle the repetitive stuff, and you can focus on the human activities. That's how you build a business that works for you instead of burning you out. So, I'll put it a link down below like I said, we're helping people just like yourself watching videos like this that maybe don't want to be a content creator. Well, let's build out a system that can help you get some more views on your product to help you start making more money. We'll see you guys in the next one. Have a blessed day. Bye-bye.
Ready to start building with AI?
Join the Shipping Skool community free. Paid members get live bootcamp calls every week and weekly-updated courses.
Join Shipping SkoolKeep reading
How to Make AI Write Like You: writing-style-skill for Claude Code
writing-style-skill is a free, open-source GitHub repo that learns your voice from your edits. You edit an AI draft, observe.py records the before and after, and improve.py pulls rules from the changes and ranks them by confidence. Those rules get written back into a SKILL.md file that works with Claude Code or OpenClaw.
She Runs 9 AI Agents That Manage Her Entire Life. Here Is Her Exact Playbook.
A startup founder scaled from one AI agent to nine, managing her entire life and two businesses. Here is her exact playbook, mistakes included.
OpenClaw Marketing Agents: A 7-Agent Stack for Solo Founders
An indie hacker named Rithik built 7 single-purpose OpenClaw agents: an orchestrator plus agents for Reddit research, writing, repurposing, tweets, trend monitoring, and engagement. He says it took his marketing from 15-20 hours a week to almost zero. My advice: start with one agent and add more over time.
