This is an AI side hustle for laid off workers, told without the vague success-story fluff. I'll get the specifics out of the way first because the internet doesn't need another one of those.
I lost my job in February. Sales coordinator at a logistics company in the Midwest. The company downsized, I was 18 months in, last hired first fired. Two weeks severance. I had about $4,200 in savings, a roommate situation that needed $700 a month from me for rent, and zero plan.
I'm writing this in late June. As of last week I'm earning enough from AI services for small businesses to cover my rent and half my groceries. Maybe $1,600 to $1,900 a month, depending on how many billable hours I put in. That's it. That's the whole income. I'm not buying a Tesla. I'm just not panicking about rent.
This post is how that happened. Some of it I figured out on purpose. A lot of it was luck and grinding.
One thing before I get into how it happened. The most useful piece of advice I would give past-me. Do not try to learn this from YouTube. I spent the first month after the layoff watching tutorials. Zapier. n8n. Make. Cold outreach. Pricing. Every video contradicted the last one. I built nothing in that month. The fastest way I ever bought back time on something this confusing was paying for one structured course and just finishing it. YouTube is fine once you already know what you are looking for. It is brutal as a starting point.
The TikTok that started it
About three weeks after the layoff I was doom-scrolling at 1am. The layoff doom-scroll, not the regular doom-scroll. And a TikTok from a guy named Liam Ottley came across my feed. If you don't know him, he runs a YouTube channel about building AI agencies for small businesses. The video was him explaining how he'd taken a Vapi voice agent (AI phone-answering bot, basically) and sold it to a chiropractor for $1,500 a month to handle their after-hours appointment calls.
He was holding up the dashboard. The numbers were real. He'd been doing this for a while and had built a whole community around it.
I had nothing else going on. I figured I had nothing to lose.
What the TikTok did NOT tell me
Here is what the TikTok told me. The business model exists, people pay for it, the math works.
Here is what the TikTok did NOT tell me. How to actually do it.
That part wasn't malicious. He has paid courses for the how-to. He's clear about that. The free content is the pitch. But for me, sitting on a couch with $4,200 and no client and no clue how Vapi worked, the gap between "this is possible" and "I can do this Tuesday" was a Grand Canyon.
The YouTube spiral that almost killed me
I did what everyone does first. I went to YouTube. "How to build AI voice agents for local businesses." "Vapi tutorial." "How to land your first AI agency client."
There are about a thousand of these videos. They are all useful and they are all useless at the same time. Useful because the information is in there somewhere. Useless because every creator has a slightly different stack, slightly different cold outreach script, slightly different pricing model, and they're all confident.
I watched probably 25 videos in five days. Every video I watched, I felt smarter for an hour and dumber for two. I had a Google doc with 60 bullet points of "tips" that contradicted each other. I still hadn't touched the actual tools.
This is the rut where most people quit. I almost quit. I had a job interview at a different logistics company the following Monday and I was halfway convinced I was going to just take it and shelve the whole AI thing.
Why I bought a course
I bought a course out of frustration with YouTube, basically. Not because I thought a course would magically work. Because I needed someone to put the steps in order so I could stop optimizing and start doing.
I looked at the main paid options. Skool had a community-based one that was $99 a month plus a $300 upfront, which felt like a lot to commit to while unemployed. Skillshare had something but the reviews said it was outdated. Coursera had a real AI for business specialization that was $49 a month but skewed academic. Felt like learning theory, not running a business. There were also creator courses in the $497 to $2,000 range, including Liam's own. I couldn't justify the spend, not on $4,200 of savings.
I ended up on a thing called Mindwand. It came up on a Reddit thread comparing AI courses for beginners. People kept saying it was structured weirdly. Short daily lessons with characters and exercises. But they all said they actually finished it, which is what caught my eye. It was around 89 bucks. Cheapest of the lot. I bought it the same night.
I was prepared to be disappointed. I was not.
The lessons take 15 to 20 minutes. They aren't lectures. They're little story-based scenarios where a character is trying to solve an AI problem and you do the steps with them. It sounds dumb on paper but the format meant I actually opened it every day. I'd opened my Coursera certificate from 2022 about three times in three years. I opened Mindwand 17 days out of the first 20.
The course didn't directly teach me how to build a Vapi voice agent. That's not what it's for. What it taught me was prompt structure, how to talk to AI without producing slop, and the discipline of using AI as a tool inside a workflow instead of as a magic answer machine. That foundation was the missing piece. Once I had it, the YouTube videos about Vapi made way more sense, because I knew how to translate what they were teaching into prompts I could actually run.
The actual grind
I built my first Vapi voice agent on a Sunday. It was for a fictional dental office I made up so I'd have something to demo. It was bad. The agent kept hallucinating insurance information. I rewrote the prompts six times over two weeks before it stopped doing that.
Then I started cold-outreaching local businesses on Instagram and Google Maps. I sent about 80 DMs and emails in the first week. Got two responses. One was "no thank you." The other was a chiropractor (same niche Liam talked about) who was curious enough to get on a call.
She did not buy. But she told me my pitch sounded too technical and she had no idea what "voice agent" even meant. I rewrote my pitch.
Sent another 60 messages the next week. Two more responses, one demo call, no sale.
This is the part of the process that nobody on YouTube films. The "you sent 200 messages and got two demo calls and zero clients" part. I almost quit during this stretch too. I posted in a Discord I'd joined about being demoralized and someone responded with "you're three weeks in, this is normal, keep going." That was the only reason I kept going.
Week 6 I landed my first client. A small dog grooming business that wanted an AI to answer the phone during salon hours when the receptionist was with a dog. $400 a month. I priced it that low because I was scared they'd say no.
Week 9 I landed my second client through a referral from the dog groomer. A photographer who wanted a chatbot on her website to handle booking inquiries. $650 a month.
Week 14, last week, I landed my third. Real estate agent. $700 a month for a Vapi agent that handles initial property inquiries after hours.
That's the math. $1,750 a month, recurring, three clients. Roughly my rent and half my groceries. Not life-changing. But I'm not panicking and the second job interview never happened.
The tip I actually believe
People keep asking me "what's the trick" and I want to be honest about this.
I can't tell you what will work and what won't. Some of the prompts I tried in week 2 are still in my main stack. Some of them I threw out by week 6. The outreach script I'm using right now is the seventh version. I don't know if it'll be the version I'm using in October.
What I can tell you is that if you stick with the thing long enough (long enough being measured in months, not weeks) you'll figure it out. The figuring-out is the prize. People who quit at week 3 quit before the figuring-out can even start.
The course gave me the structure. The grinding gave me the clients. Neither one of those alone would have worked.
What I wish I'd known on day one
Mostly that the income curve is not a curve. It's flat for two months and then it isn't. The two months feel like forever when you're in them. They look short in retrospect.
If you're reading this with a similar layoff or career-fork situation, I'm not going to tell you to do what I did. I genuinely don't know if it works for everyone. The market for AI services for local businesses is real but it's not infinite, and the next 12 months are going to thin the field of people doing it.
What I'd say is. Pick the smallest possible bet you can make. Mine was an $89 course and three weeks of trying not to give up. If your bet has to be smaller than that, find the smaller one. Just don't let the bet be zero.