4 week sprint - 0 to AI agency in 30 days

 I'm giving myself 4 weeks, 30 days, to go from a few social accounts and minimal experience to a (solo) AI agency. I'm hoping to have a functional and working agency by the end of the four weeks, including products, services, and reoccurring revenue. If you're unfamiliar with the field, or AI in general, four weeks is and isn't a lot of time to build an agency. 

    My name is Joseph, I am 24 and I've never worked for or ran an agency before. I've worked in multiple sales roles since I was 18 and that is the current field that I work in. I enjoy it, the thrill, the process, it's all something that I enjoy doing - and I'm good  at it. But it's time consuming, it's draining, and there are plenty of days I ask myself why I stay in this field of work. So why am I doing this, and why is the timeframe so short? 

    I'm doing this for a few reasons. Above all I'm interested in everything, literally everything. Some utilize the term 'Polymath' though I don't have enough belief in myself to consider myself to be titled as such. Additionally, I have a profound interest in Artificial Intelligence. The same interest people had when computers became available to the typical household and some people realized "Wait, this terminal thing let's me run the computer without clicking around?" or "If I write this code and save it as a certain file type, it will run the exact set of commands I want, automatically!" of course there is a vast difference between a .bat file and Artificial Intelligence. Lastly, I'm working towards my own agency because I strive and dream of working for myself with my own rules and under my supervision - oh, and money. 

    Why am I qualified to do this? I'd argue that I'm not. At least not yet - but that's part of the whole point of doing it publicly. If  I slack off and don't hit my goals or post my updates, people will notice and then that's on me. I've spent the last 3-4 months working on building several AI assistants, the main assistant being S.A.G.E the first (working) AI that I designed and built. laugh as you will, but the inspiration came from J.A.R.V.I.S, Tony Stark's hyper-intelligent assistant. But, I wasn't viewing J.A.R.V.I.S as a "crazy smart, access to anything with electricity, unlimited funding and tech at it's disposal, British voiced supercomputer" I viewed him for his structure and how he interacted with Tony. Extremely modular, self healing, self diagnosing, capability to adapt from his memories, remember everything, and adjust to speak to Tony exactly as Tony needed him to in any given scenario. But enough about the MCU, back to why I'm not qualified. I've never worked a tech job and I didn't go to school for computer science, but I have ripped open more electronics to understand how they work than you've probably owned in your lifetime. I have tore through the very OS of my first laptop to figure out how everything worked, and figured out .bat files can make you feel like you've entered the matrix by the time I was 11. I don't have a data engineering degree, but I am resourceful and when I have a vision of what I want to do or build - I have the ability to tear down whatever I'm working on to figure out how to make it work. That's exactly my plan for this agency. 

    Enough about me, let's get to the real stuff - the plan. Today is August 1st, 2025. by September 1st, 2025 I will (hopefully) have a working and revenue generating agency. But, there is a lot to do in four weeks - did I also mention I have a full time job, kids, pets, and other commitments/responsibilities? no? Onto the plan. 


Week 1: Build the base. 

Week 1 is all about building the foundation - skills, networking, planning, posting, digging my feet in. This is where I build out my portfolio, my blog (you found it!), Upwork profile, Fiverr profile, network and post on TikTok and discord, and double down on the AI skills I haven't gotten to yet (Cloud hosting, API billing, RAG, GenAI pricing bundles, etc.)  This is the week I lay the foundation to start building off of. That sounds a bit cliche, sure but it's the truth. 

Week 2: Packaging and Automation.

Oddly enough, this is the portion that I have the most trouble with, not the actual AI building portion. Packaging services and models together and deploying them to a business or customers. You can't exactly just 'mail' an AI. (Ok, technically you could package a small self sufficient offline first model onto a USB drive and mail it with instructions to select 'start.sh' after opening the file on your computer) I need to decide what types of models I'm going to provide, what the framework for those models will look like and how I will customize them per customer, how I plan to upkeep or maintain them in the future in case they need assistance, and automate the portions that I can. Along with pricing. I feel like most people don't know how to do this, many shy away from anything on a computer that's more complicated than downloading and opening a folder. That's why people hire AI architects, why they pay for the service. But with only a few months of experience compared to the years and years of experience a lot of people already have I feel odd charging market value (thousands) for a FAQ bot or business AI. 

Week 3: Outreach and pilot customers. 

This is where it becomes real. I can make as many TikTok's, blog posts, and AI models as I want but the moment I start messaging people and offering these services it's go time. This isn't the type of industry you can just hop into, sell a service or product and then ghost. People pay thousands for AI models and expect support for the life of the AI. It's commitment and it's responsibility. Am I capable? Absolutely, but I have to convince myself of that. Outreach and pilot customers is simple in itself. Cold outreach, free trials or discounted models with demo's and pre-sale calls. I've been doing that for years already, nothing new there. 

Week 4: Welcome to JSON Labs. 

Week four rounds out the entire sprint and it's nothing but go time now. Wrapping up pilots, polishing models, putting the finishing touches on the pitch, making sure the cloud services is set up and ready to rock, and that none of the AI's I've build have gone rogue in the last 3 weeks and are plotting against me. This is the week I push out dozens and dozens of  contacts and push my products and services to get a few customers. Which, by the way, is my determining value of if this was successful. If I can land 3 clients by week 4 I'll call it a success. Either way, the sprint is to build an agency but I have to push myself a bit further than just that. 

    So where am I at in week one and what comes next? I've built and posted a skeleton chatbot to GitHub, built this blog, posted daily to TikTok, built a twitter, built a podcast (coming soon), created my portfolio and completed both an Upwork and Fiverr profile, and planned out the sprint. Along with the multiple AI assistants I have built for myself, I plan to build almost a dozen more as practice and test theories and try out different models. As for what comes next, I am working on a content calendar to keep posts coming consistently to build a following. I am working on networking with other AI architects and builders, and bouncing ideas of a lot of people. I will also be reaching out, without expectation, to several ecommerce business owners to inquire about what they would expect out of an AI. 

    Whether I blow my goal out of the water or publicly fall on my face, I've at minimal stepped out of my comfort zone and tried something new which to me is more important than building an agency from the ground up. Follow along to see what I pull off, or don't, during the next four weeks. 

Follow me here: 

This will bring you to any of my current pages. 

First public GitHub repo

 My first (public) GitHub repo is posted!

For the second day of my 30 day AI transition sprint, more on that to come, I wanted to scaffold, build, and push a skeleton AI chatbot to GitHub. This is the first of several "template AI's" I plan to build and push to my blog, Twitter, and project portfolio. 

I have a few reason for going through this process: 

Gain experience in designing customizable/editable AI models

Build and test start.sh file packages

Publicly share my work (and feel good doing it)

This was a simple and fun skeleton AI to build, and I'm hoping that I can see others use it and show me how they turned it into their own. It's rewarding in itself to post something that others can use, even if it's something small like a skeleton CLI chatbot running Ollama. 

But with that said, I stayed up much too late attempting to get this checked off the sprint list - and there is much to do tomorrow. 

Enjoy SPARK for yourself HERE 

Lab log {7.26.25}

 What I worked on:

~ Found a free Dell E173FPb monitor (from 2004!) along with an APC Back-UPS 550 on Facebook marketplace

~ Networked with several TikTok creators regarding projects and collabs

~ Continued to work on the Dev room, bringing in my used Dell Inspiron 3668, the Dell E173FPb, some extra peripherals, and the APC Back-UPS 550 to set up a temporary work station. *more to come on this*

~ Created short form content on for the Dev room build for TikTok.  

What I learned  / noticed:

~ I dove deep into corporate battery backup designs, VA ratings, battery sizes, and it made me start to think of how I could utilize these in the Dev room. Especially considering I'm pulling power into the room. 

~ That, while an emulator style monitor is cool, I'll want to expedite the process of dual (modern) monitors. This E173FPb will stay, but as an extra peripheral or maybe a raspberry pi monitor for a simple function or display. 

~ I noticed that the VGA cable I had laying around really does it's job well. There are drawbacks, but for the 3668 and the E173FPb, there isn't much image or speed quality to 'lose'. Everything runs smooth, there are no stutters or hesitations in the image. It's not my Omen 25L by any stretch of the imagination, but it's 100% serving it's purpose. 

 Next task / project:

~ I plan to change the 3668 from Lubuntu to Mint or maybe even Pop!_OS. Though, straight Debian is tempting as well. 

~ Add the projector to the Dev room

~ Write out the first true blog post (coming soon) 

 Random insight: 

~Set up a Facebook marketplace saved search for free within 10 miles. That's exactly how I ran into the free tech today (and I know they're putting out more soon)   

Lab log {7.23.25}

  What I Worked On

~ Collaborated with several creators on TikTok, including a possible podcast, volunteer internship for new website/company, and another AI designer

~ Built a new Google ecosystem under joseph.dev.labs Gmail. This allows me to separate my personal profiles from my dev/project profile. In doing this, slightly too late, I had to migrate my blog layout to the new account. I've since deleted the previous blog entirely.

~ Changed backgrounds for both PC's at my desk, the Omen 25L for gaming now color matches red for the lock screen, home screen, and the PC's RGB lighting. My Node-Zero Hp dev PC utilizes a blue background to color match the single LED behind the power button. I currently have both PC's hooked to a single monitor with this hdmi splitter. I've attached this splitter, along with the cords, underneath my desk where the button is within reach for a easy (and very cool) method of changing which device is being displayed.

~ Began working on, designing, and cleaning the upcoming dev and content area. More to come on this in the coming week. 

~Attempted to migrate S.A.G.E to Node-Zero via git, though I failed to remember my extensive .gitignore file and it turned into a mess. I've cleared the pulled folder and will manually zip, push to a thumb drive, and upload tomorrow. a new pip freeze > requirements.txt will need to be done, and this may be a fantastic opportunity to practice designing and implementing a start.sh file to allow for a smooth download and migration of the AI to another device. 

What I Learned / Noticed

~ Pick your Email before singing up for sites or starting a blog, you'll have to redo it all later anyway.

~ I took about an hour and tested/learned about Notebooklm.com I think this will be a very useful tool, as I become more familiar and fluent with it. It's an AI powered notebook powered by Google.

~ There are a lot of creators, designers, and diyer's that are just dying to talk to others about their interests, but nobody see's their content. I find myself in that same scenario, so when small creators find one another - the conversations just flow.

Next task / project

~ Complete the S.A.G.E transition to the Node-Zero PC

~ Find/purchase remaining furniture and lighting for the dev space project. Along with a few other lights, power supplies, and a few more secrets.

Random Insight 

> “Do it tired.”