Building a $5K/Month Dog Business Using AI Agents (Step-by-Step Niche Finder Tutorial)

If you have been sitting on an online business idea for six months or more and you still haven’t picked a niche, this is the video that changes that. Alston builds an AI agent from scratch using Make.com and Airtable. That agent goes out every single Monday morning and comes back with 10 scored sub-niches in the dog space, ranked by Google Trends data, YouTube opportunity, monetization potential, and competition level. Total cost to run this is under $10 for the first three months. Total time to build it is 20 to 30 minutes.

The agent has a name: Frank. Frank does not overthink. Frank does not second-guess. Frank pulls information, scores every niche on a numeric scale, formats the results as a clean sortable table, and delivers the whole report before you finish your morning coffee. This post walks through every step Alston covers in the video so you can build Frank yourself, or follow along and swap the dog niche for any other space that interests you.

What You’ll Walk Out With

A free Airtable tracker with auto-scored niche data that updates on a schedule you control
The exact Airtable columns and formula Alston uses to calculate a final opportunity score between 0 and 10
Step-by-step instructions for connecting Make.com to the ChatGPT API for under $10
The JSON parsing step that converts raw AI output into clean, sortable rows inside Airtable
How to build and name an AI agent (Frank) using GPT-4 Turbo inside Make.com
Why micro-niches outperform broad categories every time, with a real-world income analogy
How to find your own starting niche using the free tool at finder.platformproof.com

Why Most Online Business Starters Stay Stuck

Alston opens the video with a direct observation: new online business owners struggle with three things. Speed. Emotion. Second-guessing. Most people start with a burst of motivation and then spend weeks in the research phase. They read 40 articles, watch 20 videos, and never commit to a niche because the decision feels permanent. Then the emotion kicks in: what if I pick wrong? What if the niche is too small? What if someone already does this? And that overthinking turns into paralysis.

His point is that money loves speed. The longer you spend in analysis mode, the more momentum you lose. And if you have been stuck for longer than six months, or longer than a year, you are not missing information. You are missing a system that removes the emotional weight from the decision. That is exactly what an AI agent does. It collects data on your behalf, scores the options against real metrics, and removes the guesswork. You make the final call, but you make it with numbers in front of you instead of feelings.

Why the Pet Niche Makes Sense as a Starting Point

Alston chooses the pet niche for this tutorial, and he gives four specific reasons. First, it is evergreen. People will always own pets and will always spend money on them. That does not change during recessions or seasonal shifts the way some other markets do. Second, there is high commercial intent. Pet owners are actively looking to solve problems and they are willing to pay for solutions. This is the kind of market where buyers show up ready to buy.

Third, when you zoom into specific micro-niches inside the pet space, competition drops sharply. Most people make the mistake of trying to build a page about dogs in general. That market is saturated. But a page specifically helping Doberman pincher owners, or people raising miniature schnauzers, or wiener dog owners with back problems, that is a different story entirely. Those audiences have very specific needs that general pet content never fully addresses. And fourth, this type of business does not require your face, your voice, or personal expertise as a dog trainer. You can build it as a content and affiliate operation from the start.

Why “Dog” Is Too Broad and What to Do Instead

Alston uses an analogy here that lands well. A general doctor earns around $288,000 per year. A cardiothoracic surgeon earns around $400,000 per year. Same field, very different income. The specialist commands more because they solve a specific problem at a deeper level. Your online business works the same way. When you try to talk to everyone in the dog niche, you wind up talking to no one. When you talk specifically to people whose miniature schnauzer has anxiety, you suddenly sound like the exact person they have been searching for.

The Doberman pincher owner and the wiener dog owner both love their pets, but their needs and requirements are completely different. Different body types, different health concerns, different temperaments, different product needs. If your content speaks to one of them specifically, it feels personal. If it tries to address all dog owners at once, it gets lost in a sea of generic content that already exists. Niching down is not about limiting yourself. It is about becoming the obvious choice for a smaller, more committed audience.

The Three Tools You Need to Build Frank

Alston keeps the tool stack simple. You need three things to get started:

Make.com: the automation platform that runs your scenario. You can start for free and run a few scenarios without paying anything.
Airtable: the database where Frank stores and displays the niche results. It is free to use and works like a cross between Google Sheets and a proper database.
ChatGPT API: the intelligence behind Frank’s research. You need less than $10 to get started, and that amount will run the agent for roughly three months.

Down the road, Alston mentions you can also add Google APIs, pull from Reddit, TikTok, Instagram, and forum sites to give Frank even more data sources to work with. But for the initial build, you do not need any of that. The three tools above are enough to generate genuinely useful niche research on a weekly schedule.

Step 1: Set Up Your Airtable Niche Tracker

The first thing Alston builds is the Airtable table. He goes to airtable.com, creates a new base, names it “Dog Stuff” for now (you can rename it later), and then creates a table inside called “Doggy Niches.” The table needs the following fields, set up in this exact order:

Niche Name: Single line of text. This is where Frank fills in the sub-niche label.
Google Trend Score: Number field. Frank scores each niche on Google Trends data.
YouTube Opportunity: Number field. How much content opportunity exists on YouTube for this niche.
Monetization Score: Number field. How easily this niche converts into revenue.
Competition Score: Number field. Lower competition gets a higher score here.
Final Score: Formula field. The formula adds all four numeric scores and divides by four. This gives you a clean 0 to 10 score for sorting and comparing.
Content Ideas: Long form text. Frank fills in three specific content ideas for each niche.
Monetization Plan: Long form text. A monetization idea tailored to each specific niche.
Date Created: Date field with default set to current date. This lets you track which niches were researched recently versus months ago.

One thing Alston catches partway through: the formula needs to divide by 4, not by 10. When you add four scores and divide by 4, you get a clean average. Dividing by 10 gives you compressed numbers that all cluster near the same value. He fixes this live in the video and notes that if you put bad math in, you get bad output. The same principle applies to prompts.

Step 2: Connect Make.com to ChatGPT

With the Airtable set up, Alston moves to Make.com. He creates a new scenario and starts by adding an OpenAI module. The module type he selects is “Create a Chat Completion.” This is the step where Frank receives the research prompt and returns a list of niches.

To connect the module you need a ChatGPT API key. Alston walks through where to find it: go to the OpenAI API platform, create or log into your account, and follow the steps to generate a key. Once the key is in, you select the model. Alston uses “chat for latest” so the scenario always picks up the most current ChatGPT version available. As of the video, GPT-4 Turbo is the choice for the agent itself.

The prompt is the most important part of this module. Alston writes a detailed, specific prompt that tells ChatGPT to find 10 low-competition, high-demand sub-niches for a given niche (in this case, “dog”). The prompt specifies the exact JSON output structure you want: niche name, Google trend score, YouTube score, monetization score, competition score, final score, an array of content ideas, and a monetization plan. The specificity of the prompt is what separates useful output from vague filler. He makes the point directly: bad input produces bad output.

Step 3: Parse the JSON Output

After the OpenAI module runs, the raw response needs to be cleaned up before it can enter Airtable cleanly. Alston adds a JSON parser module to the scenario. He clicks the plus sign, types “JSON,” and selects “Parse JSON.”

The parser needs a data structure that tells Make what shape the data is in. Rather than building the structure by hand field by field, Alston uses the “generate from sample” option. He pastes in a sample JSON block that matches the output format he specified in the prompt, and Make automatically creates the data structure for him. He then changes the numeric fields (Google trend score, YouTube score, monetization score, competition score, final score) from text to number type, and confirms content ideas stays as an array and monetization plan stays as text.

The JSON string input maps to the result from the OpenAI module. This is the bridge that takes messy AI output and converts it into structured rows that Airtable can receive. The scenario will show an error at this stage because there is no live data yet. That is normal. The error resolves as soon as the full scenario runs end to end.

Step 4: Connect Frank to Airtable

The final module in the Make.com scenario is an Airtable “Create a Record” module. Alston clicks the plus sign, searches for Airtable, and selects the record creation option. This step connects Make to your specific Airtable base and maps each field from the parsed JSON to the corresponding column in your table.

To connect Airtable you need an API token from your Airtable account. Alston shows where to find it: go to Airtable, open integrations, and grab your token. Once connected, Make pulls in the list of bases and tables. He selects the “Doggy Niches” table, and then maps each field manually: niche name goes to Niche Name, Google trend score maps to the Google Trend Score column, YouTube score maps to YouTube Opportunity, and so on through monetization plan. Date created populates automatically from Airtable.

Once the mapping is complete, he saves and runs the scenario once. The result: 10 new rows appear in Airtable, each with a niche name, four individual scores, a final combined score, three content ideas, and a monetization suggestion. The date populated automatically. The scenario worked end to end.

What Frank’s Weekly Reports Look Like

Here is a sample of what showed up in Airtable from the test run in the video. These are real outputs from the live session:

Dog mental stimulation games
Urban dog hacks
Raw feeding for beginners
Dog enrichment
Dog anxiety solutions
Homemade dog treats
Senior dog mobility

Each row comes with content ideas already attached. For “dog mental stimulation games,” Frank suggested: “Top three brain games to stop boredom fast.” That is a video title you could use today. The monetization idea for that same niche was: “Sell downloadable PDF guides for mental stimulation.” That is a $7 to $27 digital product you could create in an afternoon using the same AI tools.

Alston sorts the table by final score, highest to lowest, to quickly find the best opportunities at the top. The whole thing takes about 30 seconds to read and gives you a clear starting point for the week without any manual research. Dog anxiety solutions ranked highly in the test run, which makes sense given the volume of search traffic around that topic.

Creating the Actual AI Agent: Frank

The scenario in Make.com is the automation. But Alston also creates a proper AI agent inside Make, which is what turns the scenario into something that can run on its own schedule. He goes to the agents section, clicks “Create Agent,” and names the agent Frank. The model is GPT-4 Turbo. He then adds a detailed system prompt that tells Frank exactly what to do: find sub-niches, evaluate them against specific metrics, return data in the required JSON structure, and write content ideas that are specific enough to act on.

After the system prompt is saved, Alston goes to System Tools and adds the “Niche Finder” scenario as one of Frank’s available tools. This is what connects the agent to the automation. When Frank runs, he can trigger the Make scenario directly. The full name for the scenario inside Make is “Doggy Niche Finder.” The scenario is set to run on demand rather than on a fixed schedule because Frank controls the timing from the agent side.

Scheduling Frank and What Comes Next

Once the agent is built and the scenario is saved, Alston sets Frank to deliver a new report every Monday at 7 a.m. You can adjust this to any frequency you want. If you want to run it every Friday, or every two weeks, or on demand whenever you sit down to plan content, Make supports all of those options.

The next step after niche research, which Alston says will be a future video, is content idea research. Once you know which niches score highest, the next agent will go find specific content ideas within those niches, checking what is trending on YouTube, what is getting views recently, and what topics have demand but not too much competition. That agent feeds Frank’s output and turns a scored niche into a content plan. The idea is that each agent in the chain handles one job, and together they cover research, ideation, and eventually publishing.

Not sure which niche is right for your skills and situation?

Skip the guesswork and get a personalized recommendation at finder.platformproof.com.

Honest Drawbacks of This Approach

Frank gives you data, not guarantees. The scores are generated by ChatGPT based on its training data, not live scraped numbers from Google Trends or the YouTube API. In the base version of this build, the “Google Trend Score” is ChatGPT’s estimate, not an actual real-time API pull. Alston mentions you can add Google APIs and other integrations down the line, but the version in this video runs on AI estimates only.

That means you should treat the output as a shortlist for human validation, not a final answer. Frank narrows your options from hundreds of possibilities to ten candidates. Your job is to take the top three or four and spend 15 minutes manually checking them on YouTube and Google before committing. That is still a major time savings compared to starting from scratch, but it is worth knowing the scores are model-generated, not API-verified.

Second drawback: the output quality depends entirely on the quality of the prompt. Alston mentions this more than once in the video. When he initially ran the scenario without specifying a niche, the output was generic and unusable. The moment he added “dog” to the prompt, the results became specific and actionable. Writing a tight, detailed prompt is a skill, and it takes a little iteration to get right for your specific niche.

Third: this is step one of a multi-step system. Frank finds niches. A future agent finds content ideas. Another handles publishing or distribution. If you are hoping to set this up once and have it generate income without additional work, you will be disappointed. What it does do is remove the research bottleneck and help you move from stuck to in motion. The actual content creation and monetization still require your time.

Find Your X

The dog niche works as a tutorial example, but your starting point might be somewhere completely different. RVs, sourdough baking, vintage sewing machines, youth baseball coaching, apartment container gardening. Any niche with passionate buyers and specific sub-problems is a candidate. If you are not sure where to start, the free Finder tool at finder.platformproof.com matches you to a niche and a monetization approach based on the skills and interests you already have. It takes about two minutes and gives you a specific starting point instead of another list of options to overthink.

Frequently Asked Questions

Do I need coding experience to build this?

No. Alston says this directly in the video: you are essentially a programmer in the sense that you are giving instructions to a system and getting outputs back, but you are not writing code. Make.com uses a drag-and-drop module interface. Airtable is a spreadsheet-style tool. The only “technical” step is pasting in a JSON data structure and connecting an API key, and Alston walks through both of those step by step.

How much does it cost per month to run Frank?

The ChatGPT API costs less than $10 to start, and Alston estimates that amount runs the agent for roughly three months of weekly use. Make.com has a free tier that covers a limited number of scenarios and runs per month. Airtable’s free plan is sufficient for the setup shown in this video. So the real ongoing cost is a few dollars per month in OpenAI credits once you exhaust the initial amount.

Can I use this for a niche other than dogs?

Yes. The prompt Alston builds asks ChatGPT to find sub-niches within whatever niche you specify. He mentions RVs as another example. You could swap “dog” for any interest area: houseplants, pickleball, sourdough, homeschooling, vintage cars. The Airtable structure and the Make.com scenario stay the same. The only change is the niche keyword in the prompt.

Are the scores Frank returns accurate?

The base version uses ChatGPT’s internal estimates rather than live API data from Google Trends or YouTube. Alston says you can add those integrations over time, but for the starting build, treat the scores as a relative ranking rather than precise measurements. Use them to narrow your list, then spend a few minutes validating your top picks manually before committing.

How often should Frank run?

Alston schedules Frank for Monday mornings so the weekly niche report is ready when the work week starts. You can set this to any frequency Make.com supports, including daily, weekly, bi-weekly, or on demand. For most people starting out, weekly is enough. Running it daily creates more data than you can realistically act on when you are still building content in one niche.

What do I do with the niches Frank returns?

Sort by final score and look at the top three to five results. Check whether there are already YouTube channels in that space and how many subscribers they have. Search Google to see what content already exists. If the niche has search volume, some competition, but not dominant players with millions of subscribers, it is worth testing. The content ideas and monetization plan that Frank returns give you a practical starting point for your first few pieces of content.

What is the membership Alston mentions in the video?

Alston has a community where members get access to the blueprint for this exact Make.com scenario (which can be directly imported), the system prompts for Frank, and ongoing updates as he builds more agents in this series. He mentions the price was at its lowest point at the time of filming and that it would increase as more content and tools were added. He also mentions selling individual blueprints for around $7 to $10 for people who only want a specific piece.

Is the dog niche already too saturated to enter?

The broad “dog” niche is saturated. That is why this entire video exists: to help you find the specific sub-niches that are not. Doberman pincher care, miniature schnauzer anxiety, senior dog mobility support, raw feeding for small breeds. Those are all within the dog space but are specific enough that you can become a recognized resource without competing against channels with millions of subscribers. The goal is to own a micro-niche, not the whole category.

Sources

Alston Godbolt, “Building a $5K/Month Dog Business Using AI Agents,” YouTube, youtu.be/GhESM6m2kQU
Make.com automation platform: make.com
Airtable database tool: airtable.com
OpenAI API platform: platform.openai.com

Helping 1 million working adults make their first $3,000 online with the skills they already have. Alston Godbolt, Platform Proof.