Building an AI product before it was cool ($500k ARR)

Sébastien Night built an AI product early and grew it to $10k MRR. Then, he surfed the AI wave all the way to a big milestone: $500k ARR ($42k MRR).

I caught up with him to talk AI and learn how he did it. 👇

From 7 figures to $40k MRR

James: So where did it all start?

Sébastien: Two years ago, I was running a 7-figure training business and I saw how tedious it was for me to create large volumes of high-quality video. I realized this was a huge problem my clients were facing.

So I had this vision of building a tool that would solve the problem. This tool became OneTake AI. By the end of 2021 we had a proof of concept and in 2022 we had the MVP, but nobody cared about AI.

James: Too early?

Sébastien: Yes, one year too early.

James: So nothing happened?

Sébastien: I still managed to grow it to $10k MRR after a year, which is when I decided to shut down my other business. I could see that the opportunity was so much bigger in the AI B2B SaaS subscription business.

James: Looks like you weren’t wrong.

Sébastien: And the timing was great: Once ChatGPT captured people's attention, we started growing much faster. We grew 4x in 2023, from $10k MRR to 40k+ MRR.

Differentiating an AI product

James: So you were ready to surf the AI wave. But how did you differentiate your product after everyone else started building AI products?

Sébastien: We are an AI startup in the video space, but contrary to pretty much everybody else, we are not even trying to play in the "generative" field (talking avatars, etc.). Our work is purely "transformative": taking real-life recordings and editing/translating them.

James: Why?

Sébastien: I believe it’s a solid long-term choice, because even after the AI content apocalypse has submerged us in AI-generated content, there will still be people who want to record themselves (not an AI avatar) for a podcast/conference/course/etc. It is much more differentiated to be the go-to solution there.

Apocalypse now

James: Smart. Tell me more about this content apocalypse.

Sébastien: I was at a Mastermind last quarter where a successful entrepreneur was telling me all about his brand-new Zapier workflow that searches for top-ranked viral content, transcribes it in one tool, rephrases it in another tool, has a digital avatar of him generated in another, and then posts the AI-regurgitated copycat video to YouTube. This single person was aiming to churn out 400 (four hundred!) of those videos per week.

James: 😑

Sébastien: And he is just one drop of rain in the deluge.

I think "normal humans" don't want to consume this. But even more importantly, they don't want to filter and choose by themselves.

So they will abide by whatever the big tech giants want to show them.

Companies are then facing a battle, not against hundreds of blogspam sites so that you can get to Google's top 10 — that ship has sailed — but against millions of content spammers so that you get to be the single source that Bard or ChatGPT or Perplexity will spit out. Not 10-in-100 odds, but 1-in-1,000,000 odds.

James: Sounds dire.

Sébastien: I agree. We are right on the verge of a lot of convincingly realistic AI content getting poured onto every text, audio, image, and video platform.

For a short while, those platforms can dig larger moats around their "real" organic human-generated content, and it might still hold for a few months. You and I and everyone else who's playing by the "old" rules will have to go through a bajillion CAPTCHAs and proof-of-humanhood tests. After a while it will be so annoying to have increasing frustration and diminishing returns that most of us will also join the "let's automate everything" tribe.

James: So what should indie hackers do?

Sébastien: Get as much "critical mass" of human attention as you can in the next few months to then snowball and leverage it into getting more and more of those AI-picked recommendations.

Backed into an AI corner (+ tech stack)

James: Okay, let’s zoom back in. Was it just luck that your product surfed the AI wave or did you predict it?

Sébastien: I didn't build an AI startup because I wanted to do AI. I built an AI startup because I was pissed and only AI could solve the problem.

James: Why were you pissed?

Sébastien: Picture this: I'm just a few weeks into the project and we've come out with a "proof of concept" of what the product could do. I sent an email to a group of entrepreneur friends with a demo.

I was taking a very dense, rough recording from a famous speaker on a complex technical topic and in 45 minutes I turned it into a perfectly edited presentation with slides, visual transitions, key takeaways, handouts; the works. This is a project that would take an experienced high-paid editor more than 12 hours of work to even approach what I did in 45 minutes... I was so proud of my baby.

A few hours later, my friend Victor Damásio calls me on the phone to say:

“I don't want to work 45 minutes on one video. I know, I know. Normally it would be 12 hours and I should be happy. But really, what would happen if it's 12 hours of work is... I just wouldn't do it at all. I don't want 12 hours, I don't want 45 minutes. What I want is to push a single button. Can you do that?”

James: Heck of a come-to-Jesus moment.

Sébastien: I was pissed. It was impossible. It was an unreasonable request.

And that night as I turned and turned in bed, I realized that if customers have unreasonable expectations, then the first startup to make them a reality wins.

That's why I turned to AI.

I couldn't expect that ChatGPT would blow up a year later. I was prepared to face an uphill battle, to build and sell this as an AI product, simply because I could tell that AI was the way to solve my customer's impossible problem. And I knew if we could pull off "the impossible" then we would win.

James: What tech stack did you choose?

Sébastien:

• Backend: NodeJS, MongoDB, everything on DigitalOcean

• CDN: Bunny.net

• Video ingress: Uppy/Transloadit. All the video files we process are stored both on Bunny and Backblaze, so that we have at least 3 copies of any video in 3 separate locations, which is step one for offering an SLA (service level agreement) — we will soon launch a 99.9% availability guarantee to match Vimeo.

• Front-End: React + MaterialUI + the MinimalUI template

• AI: We prototype all features by assembling or repurposing open-source code or external APIs. We bend things far outside of their intended use. For example, we premiered our OneTake Chat feature by using a RAG (Retrieval-Assisted Generation) system from BasePlate.ai plus the OpenAI API. And then we hacked on top of that to build what we needed to launch. We've learned a lot from almost 6 months of daily use by hundreds of paid users, and now we're ready to transition from this to a home-built system that's tailored to what we've learned. To run our own AI code, we use GPU servers from LambdaLabs. We have a number of features like that, where users ask us for features that can't be built with current state-of-the-art models, so we build our own.

Know your customer

James: Was founder-market fit important to your success?

Sébastien: I understand the needs and psychology of the entrepreneurs we sell to better than anyone else.

This allows me to drive the product and create directly useful features that have a "Wow!" effect for customers: They couldn't have described that this was the solution they were looking for, but when they see it, they must have it.

James: That’s a neat talent.

Sébastien: Well, I've been in this space and sold to the same buyer profile for almost 15 years.

James: So it’s just a factor of time spent with one customer type?

Sébastien: That, and actually talking to people! We have a chat widget on our site and I’ve never used AI for it. Plus, the way I sell is on live webinars and live online challenges. The vast majority of paid customers come from there. And our paid customers are invited to keep joining them as well.

James: Why include paid customers?

Sébastien: So, if you've got 20 paid customers sitting in the room with you every time you talk to a new prospect, two things happen:

1. One, you get social proof and trust much faster because the lead doesn't have to trust my word that we've got good customer service. They can literally ask everybody else on the call.

2. Two, if there's currently a big breaking bug or a major annoyance in the app, customers will tell you, right there while they have you "within arm's reach". So we’d better listen and get things working smoothly ASAP, all the time, otherwise every single webinar will be dead on arrival.

James: So, to your second point, you’re kind of forcing yourselves to deliver.

Sébastien: We built a system where customer feedback is impossible to ignore. It's a kind of pressure that I think makes it impossible to fail.

Find a niche (and stay there for 15 years)

James: Do you think it’s important for indie hackers to stick to one niche like you did?

Sébastien: The legendary CEO mentor, Annie Pratt, says every entrepreneur's career is rolling a boulder. But what most of us ignore is that you can decide if you want to roll the boulder uphill or downhill.

James: So that’s a “yes”?

Sébastien: Here's how you pick your niche:

1. Pick a project that you're built for. Something that matches your strengths, weaknesses, passion, motivation and ambitions. That's like picking a boulder that is the right size for you and "feels right".

2. Pick a business model and execution that has momentum. This means (a) choosing an industry that is on fire, (b) selling a product or service that people are already spending money on, and (c) having a strategy that is already proven, high-leverage, sustainable, and easy. That’s like picking a boulder that rolls easily.

3. Be in the right place at the right time. Each market and industry has its "perfect moment in time". You want to get started in one of those. Where is the growth? Where will the growth be and remain for 10 years? (Picking a starting spot where the ground slopes sharply down, instead of being flat or starting at the foot of the hill looking up)

If you've got all three, then the signs will be there very fast. You'll be picking up momentum. In that case it makes sense to stay in one niche and dominate.

James: Like you did for 15 years.

Sébastien: But if you feel like you're still fighting your way uphill, despite giving it your all for a year or more, then maybe you should reassess.

Focus on one simple thing (MRR)

James: Okay, so you’ve got your niche and you know your customers. Then what?

Sébastien: About six months into the project, my friend David Albert (a brilliant engineer himself) asked me why the hell my MRR was still basically flat.

I gave all sorts of answers about how we hadn't reached product-market fit yet, and our webinars didn't convert yet because of missing features X and Y, and so on.

James: What did he say to that?

Sébastien: He made me see that all of these were just really smart excuses.

James: Lol.

Sébastien: As soon as I started focusing on one simple thing — "How can we grow our MRR +10% or +20% compared to last month?" — everything changed.

James: How so?

Sébastien: You can't say, "I'll just keep making the product better, and more people will come". That doesn't work.

But somehow, "I need more people to come and I'll focus on that" will drive you to make a better product: You'll have to make it better in the eyes of the buyer. And that’s what counts.

James: So what changed?

Sébastien: Everything was more clear. It was clear which features to build. It was also clear which team members to hire and keep. And it was clear how to talk about the product better.

James: What types of features were clearly going to push the needle?

Sébastien:

1. Features that solve buying objections

2. Features that onboard new users and make their lives easier (which increases activation and thus LTV)

3. Features that give better results, easier, to activated users and turn them into "power users"

4. Features that keep "power users" happy

In that order.

Top 3 growth tactics

James: So your strategy and roadmap shifted, thanks to that conversation. Other than that, what growth tactics pushed the needle most.

Sébastien: Three main tactics have helped:

• My own opt-in email list of 60,000 entrepreneurs in French-speaking countries,

• We pay affiliates a 20% lifetime commission. We run Facebook Lives and Zoom webinars to test audiences of influential online entrepreneurs: Roughly 33% of our MRR now comes from English, Italian, and Japanese-speaking customers thanks to this strategy.

• Starting last April we experimented every month with ad spend to acquire leads on a "4-day free challenge". We’ve brought our CAC down to $1,267 so far, which is a payback period of about 12 months. We are running new experiments with a free trial to slice CAC to a third of this. I think running the ads to a free challenge is not the best. Sending people directly to the webinar is likely to convert much better, which is our big new experiment.

We also worked a lot on reducing churn.

A simple churn hack

James: Tell me more about that.

Sébastien: Churn was one of my biggest concerns when I started. I knew I would have to budget a 20% month-on-month churn at the beginning, as that’s typical for new software products.

Obviously, I needed to get this down a lot for the business to grow.

James: And did you?

Sébastien: We've worked hard on it and got churn from 15% in the early months, to around 5% now. LTV increased from $717 to $2,079. Our average retention is 20 months.

James: How did you decrease churn so dramatically?

Sébastien: We achieved this with a strong onboarding flow and a huge activation incentive: If a new user onboards and publishes their 1st project within 30 days, we upgrade their account to "unlimited". People who have uploaded content tend to stick much longer.

As a result, 77% of our users create and publish their first video project within 30 days of subscribing.

James: I like how simple that is.

Sébastien: We also run 2-hour live sessions on Zoom bi-weekly with customers (one session in French, one in English) to showcase updates, answer questions and take feature requests.  20% of all customers attend every session.

Charge from the beginning

James: Anything else that helped you grow to $500k ARR?

Sébastien: In the startup/indie world, there is a huge belief that the easiest/fastest way to get started is to offer the product for free to get feedback, and iterate until you feel confident charging.

I think that's a convenient excuse to be scared of charging for your value, and a recipe for failure.

James: Fair.

Sébastien: We started with 2 people pre-paying for our software, 3 months before the MVP was open.

It's not hard to sell two people on a dream solution to their current problems. It's definitely not harder than getting, say, 20 random people to "give it a try for free". But those two paying customers will give you just as much, probably more, relevant quality feedback than the 20 tire-kickers. And they give you money!

James: For sure.

Sébastien: The responsibility of making paying customers happy; that's such a motivator.

Non-buyers will give you a ton of dumb, same-thing-they've-seen-elsewhere ideas. That's not what you want.

James: So what are your thoughts on the Freemium model?

Sébastien: I think "freemium" is a late-stage strategy (we're just experimenting with it), not a getting-started strategy.

When you're getting started, you need cash. And you need feedback from people who have said, with their credit card, that solving this problem matters to them.

That's why for more than a year, I ignored everybody who complained that we didn't have a free trial.

It’s easier than you’re making it

James: Alright. Any parting words?

Sébastien: I think business is simple, and I think life is simple. But most of us entrepreneurs make it way more complicated than it needs to be.

James: Why?

Sébastien: Because it makes us feel important.

James: Interesting, tell me more.

Sébastien: In every business I've run, there was a simple recipe to figure out. Simple levers you can pull everyday to make it grow with little effort. I call them "success factors". The whole game is figuring those out, and then you just focus on the success factors.

James: How do you find them?

Sébastien:

1. Find success. Who has succeeded by an outlandishly large margin after starting from similar circumstances?

2. Find what these people did along the way — not what they are doing now that seems successful. You can’t learn from what these people are doing now. This is where a site like Indie Hackers is so useful, because you can see the trajectory of the people whose success is dazzling you today. These 7-figure-plus startups might be looking down at you from up in the clouds, but the founder had to walk all the way up a creaky wooden staircase to get there. Where's the staircase? Which steps are rotten? Find their blog or their early interviews to figure out how they were thinking back then.

3. Third, learn from their mistakes. Just because someone succeeded by shedding a ton of blood, sweat, and tears doesn't mean you need to shed the same amount of bodily fluids to get there, so why not spare yourself some of the pain? If you listen to interviews and podcasts, you will soon find out that there are common traps. The common traps are easy to fall into because people fall into them when they do the intuitive thing that doesn't work. So you can easily avoid them if you're willing to be coachable and do counter-intuitive stuff.

James: That’s it?

Sébastien: You have to let go of the ego. And that's hard. Because when you start your business, nobody cares. Nobody cares about the product, nobody cares about your idea, nobody cares about the problem you're solving. Because you don't yet know how to explain it.

So then entrepreneurs try all kinds of things to make themselves significant. It's all about how they're hustling the most, or struggling the most, or getting the most confused, or fundraising the most, or getting the most press, or getting the most awards.

I've been there, and it sucks. Because none of those things matter.

James: What does matter?

Sébastien: Today, I just focus on the simple things. Getting my MRR to grow every month, every week, every day. Fine-tuning my own "success factors" for this little business.

Outside of me, basically nobody cares. And that is perfectly fine. :)

James: In case someone does care, where can they find you?

Sébastien: Check out OneTake AI. We’re also on Facebook and IH.