Quitting his job with no plan and hitting $14.5k MRR with a hybrid agency

When ChatGPT dropped, Adithyan Ilangovan quit his job with no plan. He built a product and failed. Then, he built a tech-enabled service called AI Podcasting and hit $14.5k MRR.

He did that via a hybrid model that he says is underutilized.

Here's Adithyan with the details. 👇

Jumping on the AI bandwagon

I did my master's in Video Engineering. After my studies, I worked and built video engineering infrastructure at a couple of startups and then led product at Paramount Studios. So you could say I'm a research engineer with a product background.

I wasn't super happy with thecorporate grind, and I had always wanted to start a business. It was always on the back of my mind. I wanted to do something meaningful and have some more free time for myself — ha! What a trap that was!

Anyway, ChatGPT happened, and I knew immediately that something big was coming. Things were going to change.

So, like many others, I jumped on the AI bandwagon. I quit my job with no specific plans and struggled for the first six months.

I started by building a great technical product, but it was not a good business.

Eventually, I figured stuff out. I mean, it's still a struggle, but I have some revenue now.

Building a $14.5k MRR niche agency

I run AI Podcasting (AIP), a podcast post-production agency.

Podcasters record their content, and we handle everything after that: editing, audio, notes, social, publishing, distribution, paid/organic promotion, and more.

Basically, our pitch is that you can focus on telling the story you want to tell to your audience, and we will help you deal with all the stuff that comes after clicking "finish recording."

We're at about $14.5k MRR.

Starting manually

We figured out which market to serve, then did a lot of work manually — we didn't build a product right away. We did the same things the creator or the podcaster would have done after recording their content.

Basically, we lived through their pain. And it was brutal.

We kept doing it for two to three months until we slowly started realizing what parts could be easily automated with AI and other tools.

Eventually, we were able to automate about 80% with AI. So that is basically our margin.

Our internal tool stack is:

• Python for backend

• NextJS for frontend

• OpenAI APIs for AI workflows.

• FFmpeg for media

• MongoDB

• Azure Infrastructure

• Modal for a lot of scale for Python containers

• Lots of custom automations

A hybrid model full of indie hacking opportunities

We use a done-for-you subscription model. Flat monthly pricing, usually $2.5k to $3k for 4 episodes per month, with add-on hours as needed. Revenue growth comes from expanding workflows with existing customers.

I think this is an important note: Not many people are exploring the hybrid model of heavy automation plus human polish. There is definitely a market premium and demand for that.

Everything that can be 100% automated will mean competition — and soon, it'll get sucked up by the frontier model providers.

But this hybrid model of running like an agency creates a moat. Especially with the personal relationships that you will build with the clients — and trying to understand their preferences and pains. It's hard to get in. But once you are in, you can lock in and keep increasing your margins with AI.

Things are moving rapidly, so I can't say for sure — but that's my take for now.

Organic growth

As far as growth, we could do much more — and much better.

We've grown mostly from referrals. Strong outcomes drive word of mouth. We also grow via inbound from the creator network.

We haven't done any paid ads so far.

If he could turn back time 🎵

If I had to startover, here's what I'd do:

• I wouldn't start with B2C as I did with my original project. I'd start with B2B.

• I wouldn't build a thing until I had a customer, or at least clearly understood the pain point (via The Mom Test).

• I would charge early and charge high.

If you're going to overbuild, don't bother building

If you are a hacker, most likely you will love building; that's the easy part. But you need to get out and talk to real customers early and often.

Do not overbuild, or I would go as far as saying it's better not to build anything at all at the start.

And try to understand the market very clearly. If you think you already have a good understanding, stress-test your assumptions. The Mom Test is a very good book in that regard.

If I had done this properly, it would have saved me maybe six months of pain.

Here's a very good test: "Is this person willing to give their credit card information for my solution to the problem they say they have?" That's it. Simple.

And after that, if people are willing to stick around, you're on the right track.

What's next?

I want to scale the production engine while keeping quality high. Serve more high-value creators. Keep improving automation. And eventually, build a tooling layer for smaller creators.

Also, I am terrible at marketing and sales. I'd like to improve.

You can learn more at AIP. And you can find me on X, LinkedIn, Github, and my personal website.