Making $14,000 in MRR with a custom ChatGPT tool
I interviewed Mike Heap and Alex Rainey, the founders of My AskAI, a SaaS that allows you to create ChatGPT with your own data.
Let's dive in!
What’s your name, and what are you working on?
We are Mike Heap and Alex Rainey.
We’re working on My AskAI, a way for you to create your own custom ChatGPT that uses your own data to answer questions, all without code.
Once created, you can share or integrate it pretty much wherever you want using our embeds, APIs, and plugins. The result is that you and your team, or your users, can get answers to questions on your content wherever you are.
What inspired you to create My AskAI?
We had already launched an AI fine-tuning product as part of a hackathon for Ben’s Bites at the end of last year and had got a bit of traction with that, but found the learning curve for most to be a bit too steep.
We had also seen people making lots of side projects where they were creating “chat with this book/podcast”-type products and thought it would be cool to make a platform that would allow anyone to do it for whatever content they wanted.
At what point did you decide to devote more time to My AskAI vs. getting back to your job?
At the time I was looking for freelance work and Alex was exploring options having not long since sold his last company. After launching our fine-tuning product we then gave an option to purchase a deal on a new product (My AskAI) we were working on.
We got around $5k in pre-sales for the product prior to launching based on just a short description of what it would do. So we knew there was demand there.
We also thought that the AI hype train had already left the station, and we didn’t want to miss our opportunity to build some cool stuff.
So at the end of January we both decided not to look for other work and go all in focusing on My AskAI with a deadline of trying to hit ramen profitability by the end of May.
How did you validate that there’s a need for My AskAI?
Firstly, when we were helping people with AI fine-tuning products, we were getting a lot of requests to be able to create "query" bots, or ways to automate customer service.
While fine-tuning could do that, it was not the optimal solution. That’s when we decided to put up a pre-launch purchase product option on our site to see how many people would pay before we built.
We put up a short bit of copy and managed to get just over 50 people to sign up to a pre-launch at $99, which felt solid given it was around 2-3x our revenue for the product we had already built and had existed!
When did you start seeing traction?
We were lucky and saw it pretty much straight away.
From the pre-launch (thank you AI hype) and the users of the fine-tuning product, we were able to work the mailing list to generate pre-sales. Then we piggybacked on that for a Product Hunt launch, which went pretty well too.
How are you currently acquiring customers?
We’re currently trying a bunch of different things including newsletter ads, we got some initial users through Twitter and LinkedIn, and try wherever possible to implement product-led growth initiatives.
But right now our biggest single source of sign ups are word of mouth referrals, which is pretty nice as it means you are building something people would actually talk about and share with friends. Our biggest channel is actually organic through Google.
Tweeting helped initially, but Product Hunt has a reliable acquisition channel so we try and launch something there once per month. Newsletters aligned to AI also helped.
Was My AskAI the first successful AI project? Did you have any (un)successful attempts in building an API product?
We started off by building a University Personal Statement writing tool that was built with fine-tuning back in October and November of last year. We got it out and got a few paying users, but OpenAI actually told us to shut it down as it went against their ethics. (We still refute this, but it wasn’t a battle we thought was worth pursuing.)
We played around with a few ideas after that before building our No Code AI Model Builter fine-tuning product and launching that at the end of last year, that got some traction, we made around $3,500 in sales in a few months, but have now sold that so we can wholly focus on My AskAI.
How is My AskAI doing now in terms of revenue/paying users?
We’re at around $14k MRR, and just over 300 users right now.
good for you, enjoy it when it lasts, there is no practical difference between what is currently going on with the so-called AI and the NFT's 2 years ago
I’d probably disagree, AI has practical use cases whereas NFTs were always looking for them.
Sure there will be a load of startups that fizzle out eventually but the core tech shift is here to stay
I think there is. AI will become commodized. NFTs just died (no useful value).
Definitely believe AI will be commoditised. Think it brings down barriers to almost every tech development, and therefore puts more focus on distribution and understanding users
it may not, in general there is a lot of value in the blockchain, but it god highjacked by the crypto and NFT, same thing is happening with "AI", a stupid amount of hype that moves the focus from actually useful to "let's slap an AI sticker on anything and sell it to the masses", case and point, someone made 14k sailing pointless subscriptions that have no value to anyone beyond the "AI" label, same as a few years ago when people both NFT's thinking they have value somehow.
Haha bit strong! Take your point but we speak to our users daily and know and can see they are using it within their businesses for a whole load of use cases, do we know the killer one? Probably not yet but certainly fun finding out!
Also believe there is still excellent potential for blockchain personally but just hasn’t been applied effectively in the vast number of circumstances yet
Chat GPT is just a text generator, whatever your customers are using it outside of "just for fun" it will fail, plus they will jump to the next "thing" as soon as it shows up, this is the way of the hype, you can only survive at the edge of it.
There are many useful things you can do with machine learning, but they require a lot of classified data and smart ways of implementing it, the hype things like chat GPT and image generators are toys, the only value they bring is entertainment. Unfortunately, there are companies that obscure that fact in the pursuit of profits and they can do a lot and very real damage (you know, people blindly trusting the output of these algorithms and acting on it) in the world, try not to be one of those.
I have been reading this thread and your messages. I think you do need to understand the technology to some good extent before commenting that "the hype things like chat GPT and image generators are toys, the only value they bring is entertainment". If you think all an LLM does is predict the next word by looking at a few before, you know only 1/4th of the story here. No customer, especially B2B customer is foolish enough to buy a SaaS toy for fun.
Look, if this is coming from a bit of ignorance (we all have some of it somewhere), you can always fill that in with a positive mindset. If it is coming from subtle envy, I do empathize with you, but you can still trust your skills and attempt to build stuff yourself to get over it. If you don't want to do AI, don't do it, build what you have faith in. But please don't spread hate towards some hardworking founder by saying "someone made 14k sailing pointless subscriptions".
you obviously have no experience with big companies, they will buy anything that will allow them to say they use the latest tech (especially if there is hype around it) and they will do it without any care if it is working, the person that does the purchase wants a demo and list of "requirements" to put a check against them, and the bigger the company is the more ridiculous the process is :)
as far as my posts here, the only thing I do here is help, and all my posts are towards that goal, some may sound harsh but there is a lot of "follow the hype" behavior here (there are some plain scams) and a realistic voice is needed here and there.
As for my understanding of Machine learning, I can't say I'm a specialist in any shape or form, but I know enough to know how much I do not know, and unfortunately, seems that understanding of Machine Learning here starts and ends with a bunch of marketing materials and API documentation.
I won't go in any argument with you because you are a "know it all" guy and we all indie hackers are scam 😂. If you actually knew anything about big companies you would have known how knowledgebases are a challenge for them, which is where this SaaS helps with. And if this is the kind of "help" you do (which IMO is envious whining), then all the more reason to conclude my discussion with you here.
But for others reading this, here is why GenAI hype is different from NFT/Crypto. Notice, that NFT/Crypto companies didn't make revenues, but got investments. They got investors and were in the Valuation game while with GenAI a lot of folks are getting real customers and playing the value creation game.
Don't agreed, I believe there is practical difference because it solves real life problems. If it didn't, then you would be right.
Exactly Sam!
We definitely have a bunch of early adopters we know will churn because they are looking for use cases but we also have people who get it and are already using it effectively
you think the above mentioned "SaaS" actually solves a real problem :)
the only reason anyone purchased a subscription was because there was a "chat GPT" attached to it, despite that there is zero added value for anyone that actually decided to use it, give it another few months for the hype to settle down, that business will disappear as real people with real problems start actually using that "service" and start getting angry when the thing spits out nonsense.
Did we utilise ChatGPT hype? Yea, absolutely, we’d be silly not to.
Does it give us access to a great user base that we can build for? Also yes!
Our product won’t be for everyone but nor is it meant to be, the people it is for really really like it, and those it’s not for, we’ll that’s fine
I do not think that there is zero added value. Some of people will leave, they are using it beacuse of the hype, it makes them happy. But some people will stay since it's useful. They'll find a usecase for themselves. That's the problem of users at some point, not the builders.
That's the key of indie hacking. Find an idea, build it, make profit, exit or die.
that mentality of "hit and run" is described very accurately what is wrong with 99% of the people on this site.
The "Throw a bunch of things at the wall and see what sticks" policy is not how you create something meaningful.
You observe, investigate, analyze, learn, plan, prepare, and execute to do that
"You observe, investigate, analyze, learn, plan, prepare, and execute" is for so-called perfectionists. Perfectionists whine, procrastinate, demotivate others, and think they are the only ones right and the rest 99% are wrong.
One size does not fit all.
Definitely one way of doing things, depends what your goals are. We’ve have tried both ways and we prefer building and seeing
Couldn’t have said it better myself Samet!
What percentage of your MRR is spent on OpenAI related costs?
Less than 10% I’d say
That's not too bad. I am building LogSpend, a tool to gain complete visibility and control over OpenAI costs with detailed reports, comprehensive dashboards, and proactive cost anomaly alerts. It could be useful as you grow https://logspend.com
Thanks for sharing, will take a look, tbh the number one feature I see that is useful to us is the anomaly alert - and to be able to identify where the spike came from exactly so we can react
The tool developed by ChatGPT, an advanced language model, brings numerous benefits to its users. With its natural language processing capabilities, it assists in generating human-like responses, providing valuable information, and assisting with various tasks. The tool's ability to understand context, provide accurate answers, and engage in meaningful conversations makes it valuable in enhancing productivity, automating processes, and enabling effective communication in various domains. Good stuff
lol did chatGPT write this by any chance?
What's your Tech stack?
Bubble front end, Python back end
Good stuff! Does the API cost become a challenge for you as more users try out the product?
For us it hasn’t really been an issue, probably takes <10% of our revenue I’d say?
Great margins then! I'd love to sometime chat about how you're monitoring for latency, reliability, usage, rate-limiting and more. Building for that space, so speaking to companies now
Sure, happy to chat, DM me on twitter @mike_heap_
done!
Thanks for the great post
Thanks for reading!
Thank you!
Are you able to expand upon 'Newsletters aligned to AI also helped'? Which newsletters / how did you get your product featured?
We got lucky and a number featured us for free early on just because the tech was new then we have started looking into some paid ones recently, like Superhuman, Passionfroot is good for finding them
good
Thanks!
Great to know that you are doing good in a competitive field.Keep it up
Yeah think we were lucky to get in early and have benefited from some good press as a result, also try to make ourselves more visible as people behind the product which helps give confidence to businesses is our belief
I am building something similar, an open-source version of this- and there are 100s of Indie Devs building this product. Do you think you will be able to differentiate your product, and what do you feel about the competition?
Hi Mehmet.
Is your product published? Any link to try it?
I think there are some really great products out there that are smashing it from other Indie hackers eg Bhanu and Yasser.
I think a lot of people will build the core functionality but not everyone will make it easy, find a way to get it into people’s hands or the right niches or build to what user’s actually want/need.
Glib example but lots of people can build a todo list app but some people still pay for them.
All comes down to the specifics of solving the problem for a certain group
How much does Open AI API cost?
It’s not actually too bad, maybe $600/mo or so atm?
This is great! Definitely, a good position to be a platform during the hype.
Yeah, thanks! Think we definitely benefited to some extent from a first mover advantage, we were probably one of the first 3 to do this as far as I am aware
That's wonderful news! How exactly does your user make use of it? I'm curious about it.
How much does the API cost, and can it be maid more efficient to save costs?
nice man! how is it build?
Amazing
Thanks!
Thank you!
Thanks for the write up Darko, appreciate it!
Thank you for your time ^_^