[Journey] Building a 10k OFM + AI OFM active users CRM + AI Chatbot

Fellow blackhatters,
some time ago, about a year ago, I've started sharing with you my AI OFM journey. I've started it as I thought it would fit good with my competences, but I knew it was a starting point for something different.

When I was working at it, and learning for it, especially from you, I always kept a eye to see if there was any place for a solution that could help people doing that kind of job.
So I've been learning all the SFW and NSFW AI generation techniques, learned to do the marketing, did my first model, launched, learned, got my meta accounts canceled, twice, suffered a lot, you know the drill.
Once I started having some followers and a lot of chats, it was time to find chatters. Tried one, tried three, one was rude, the other disappeared without saying anything. I knew I had found a need. I needed a solution, looked for it, and the market had zero solution, at least for Fanvue.

I've started contacting my network of ex business partners and ex colleagues in order to find anyone that had skills or competences for building up a software that could take care of chats, world is flooded with chatbots, we can build one that can handle chats.

Talked with a number of people until Mirko, a former schoolmate of my brother, Electronic Engineer, told me that their other former school friend, was an investor of one of the most advanced Italian AI companies, that had one of the most talented academic level AI engineers.

Got in contact with him, we had a meeting. After one week another meeting with their CEO and CTO. Then another meeting with all their partners and investors.

It is now November 2025, and more or less in the middle of the month one company that was providing OF AI chatting launched for Fanvue as well. I was probably the first customer of their, as I setup everything in few minutes after platform availability.
My model did about 1000$ in 2 days. And their software was REALLY limited, it only would sell through scripts. There was no instructions on how to do them, I didn't have the right content, had to study and work a lot to have it ready and working.

Shared and showed these results with this AI company and their fellow development company, we all agreed it was a good idea and we had to go for it.
Funding was the next issue, we needed to find the money to develop it.
We build up an investor deck, made some very raw calculations about how much money we would have needed (and we made it wrong, since now I need I am securing more as our initial plans were far away from the harsh development reality ).
We got as partners some money in and showed it to a couple of investors and in 2 weeks we had all the money we were looking for, also had to leave out some interested investors, It was the 31st December 2025.

Developments started one week after, about january 2026. We had 3 streams of work:
- AI (the actual chatbot, then we had to develop other things as well)
- CRM (the software that handles all the conversations and let you setup your AI Chatbot)
- Marketing (we didn't have a name, nor communication message, anything)

My background is in Sales&Marketing, so the third part was easier for me. While product wise I had to transmit my vision about the product to a bunch of very smart people with a lot of experience but that never saw Onlyfans or Fanvue. So I found myself being CEO, CTO, product owner, sales and marketing manager.

That's all for part I, will jump to the first induction meeting of one of my new developers that is starting this week. Let me know if you find this journey interesting and would like me to write more about it!

let me give you a feedback i generated millions of $ in this industry, no ai chat can replace a human chatter, we all tried it, there is hundreds of solutions, but fast u realise it still cant be reproduced what humans can do, but still i hope u build a amazing crm I would be the first one to use it.

Sure write more about it. But I'm lost, is the product already working or still in development?

Anyway, enjoy the journey.

apiguy said:

Sure write more about it. But I'm lost, is the product already working or still in development?

Anyway, enjoy the journey.

Click to expand...

it's still under development! We are targeting August 26, maybe September 26 for launch!

Interesting journey. One thing I'd suggest is to document as much as possible while building, because those early mistakes and pivots often become the most valuable lessons later on. Looking forward to reading the next part and seeing how the product evolved once the dev team came onboard.

Welldone!

Interesting journey!
Looking forward to reading the next part.

Thanks guys for your interest. I will be happily keep on writing about this journey.
At the very beginning we had to start training our LLM, we woundn't use anything that was ready, so our AI Engineers started collecting base models that could be stripped, trained and reworked. I probably don't use the correct terms, but I hope I am giving the idea.
We ordered the GPUs for the various AI parts but there was global shortage so we had some issues and had to start the tests with small models, waiting for the Blackwells to arrive. (Spoiler: we had them in early may and we could switch to much bigger and smarter models).
For training I've purchesed tens of thousands of conversations of real chatters, profitable ones, where there actually was a sale, from different creators and chatter teams.

The CRM part, our main software, started development on Fanvue, that has the APIs as some of you may know.
We draw the main functionalities, got inpired by our competitor on what we liked or not, and then went with our set of features.

Most of the job in the beginning for me was trying to transmit the OFM culture to my devs and the AI engineers as they were not familiar at all with the enviroment, but they actually had fun learning it.
My head of AI Engineering, mister Marco, a total genius with dozens of publications, could surprise me a number of times when he already knew how the things should have been done or how the conversations should have been engineered.

Marketing wise my marketing and UX team came up with 3 concepts about the logo and the font, we chose one and it's the same we're using for the launch.
The look and feel has been kept for all the website development, that was completed about 10 days ago (we launched it I thing 10-11 June) while for the CRM itself we decided to adopt dark mode only so that our users wouldn't have problems using it for a long time.

We then started working at the profiles, we decided we would have 4 kinds of profiles:
- Agency profile (invites and manages creators, chatters, managers)
- Creator profile (main profile, can invite chatters and managers)
- Manager profile (can invite and manage chatters)
- Chatter profile (can chat on one or more creator, if part of an agency)
then we started working at the invite mechanism on the CRM.

Once that was finished, we could already make our first tests on our LLMs, we tested 12 before chooseing 2 as short list and then decide for one that is was the one that was performing better.
We were a long way to be ready but it was already something "real" that we could test, try and use.

This was a long one, drop me any questions and let me know if you're interested in part 3!

I'm not saying reveal all your cards but since you said that you're good at marketing, could you share some tips about marketing?

Avarage12 said:

I'm not saying reveal all your cards but since you said that you're good at marketing, could you share some tips about marketing?

Click to expand...

Well that depends from business to business. What are you willing to market? A service, a restaurant? a product, yourself?

For any business my usual process is the following: when you're thinking about a business, find a feature that is unique, that distingueshes yourself from the competition.
When you have that idea, try to think about your target: who would be looking for that unique feature?
Then when you have that: how this target people usually finds and consumes information? What are the channels they use to find information or solutions?
Once you know that you ask yourself: what kind of message would this people, in that channel, would grab their interest?
Once you get the message, you think about what would be the pillars of communications you're gonna use:
Imagine that you're selling an ice cream to people in a hot country that never tried an ice cream, on a billboard.
Your pillars could be: it's fresh, it's sweet, it's got you favourite flavour.
Those pillars would take you to the main tagline: A sweet chill you’ve never felt.

It's just an example but I hope you got the idea.
Few years ago I've opened a Disco on an island in the Mediterrean. It had a pizza shop, but didn't have a kitchen. Pizza is low level in my country, so you have to sell it cheap otherwise it would be badly perceived. I though and thought to find a solution and I've invented a new product, I called it "Zamaki" as for pizZA uraMAKI.
It was a roll of pasta filled with normal/sushi ingredients, cut like uramakis and served with chopsticks. Well I could sell a pizza for 8-12 euros, but those went for 18-22. It was the exact same thing, with the exact same cost, but it was different. I've make a campaign about it, targeting 40+ that had the budget and curiosity for that, got newspaper articles about it and my pizzaplace was always full of people that wanted to try the famous Zamaki and pay double for the exact same thing.
Then after the season (it was a summer place) the walls owner would not renew the sewage system that was needed to keep the license, so we sued it, but that's another story!

Hope everything goes smoothly. Keep us posted!

PART III

We had the profiles on, we had the LLM ready for testing.
We started deciding what kinds of functionalities we wanted to have at launch, and what to exclude to be ready at launch.
We could see our competitors, that added a lot of functionalities in their CRM, but the AI was lacking a bit. My model sold a lot for a couple of weeks and then stopped completely. My competitors tried to replicate on platform the same functionalities that you can on Onlyfans or Fanvue, and this didn't make a lot of sense to me. If you want to send a Mass Message, you can always to it in on platform. Why don't you do something BETTER than Mass Message?
We went for an approach that is: we gonna launch the best AI chatbot and have only the CRM functionalities that could make the AI work good.
We then drawn UI/UX of all these parts of the CRM and started developing them. Actually still we have some functionalities that are there to be finished, we're working at them.
AI wise we deployed a web enviroment where I could test the AI and we could start seeing how good or bad the bot was. Also the AI engineers created a system to have thousands of sintetic conversations between fan personas and the chatbot, so that another AI model could evalutate them and learn from them.
Eventually also Blackwells arrived a couple months ago, so we could switch from a smaller model that was working already to a bigger one, that gives you more reasoning power and also multilanguage capabilities.
It arrived the time when we had to reverse engineer the OF apis. It looked very hard for the dev team. Buying from external providers was a no go for me as I wouldn't want any customer to find an Out of Service timeout because the russian hacker that is providing that service went to holiday for a week. (one of my competitors was offline for 7 days this Christmas, gave problems to a lot of creators and agencies). One of my partners pointed me to a very very talented developer living in the UK. We hired him, he reverse engineered all the streams that we needed, prepared all the documentation. His job was absolutely amazing, so we hired him. It's our CTO now. And that was just the beginning: in may (2026) I understood that we had to internalize developments, as the company that was doing that was doing it good, but it felt for them a project, not a real startup.
So I've looked for senior devs, we found them, it was very hard: don't know why it is too hard to find senior devs, they told us "with AI many will be jobless...". Well it's been hard, but we hired them. Now I have an internal dedicated team working just at my startup!
We started improving the AI chat also. We had to be able to understant at what point the chat is with fan: is he horny? is he in a romantic state? is he at home? etc. But that was just the beginning. We can now identify multiple kinds of personalities and multiple situations. The chatbot handles all the matrix of possibilites and answers and chats accordingly. If you're a money slave, we'll know that and hammer you with high values contents. If you're a romantic, we're keeping the relashionship phase long as much as you need, and when things turn heated, we'll drop you content. The chatbot is capable of understanding what kind of fan it has in front and what kind of conversation is having, to adapt to it and to convert it.

Let me know if there's anything you like me to talk about in part IV!

universalofm said:

let me give you a feedback i generated millions of $ in this industry, no ai chat can replace a human chatter, we all tried it, there is hundreds of solutions, but fast u realise it still cant be reproduced what humans can do, but still i hope u build a amazing crm I would be the first one to use it.

Click to expand...

I am really honored to have you reading my journey! Looking forward to have you try the CRM and the AI chatbot, you coulb be sure give a very competent and useful feedback!

ads studio said:

Interesting journey. One thing I'd suggest is to document as much as possible while building, because those early mistakes and pivots often become the most valuable lessons later on. Looking forward to reading the next part and seeing how the product evolved once the dev team came onboard.

Welldone!

Click to expand...

I've touched the dev team onboarding today!

Eldiablo80 said:

I am really honored to have you reading my journey! Looking forward to have you try the CRM and the AI chatbot, you coulb be sure give a very competent and useful feedback!

Click to expand...

Nothing but support here, if u would make a good solution for this u would make some serious $
Keep up with the work and your project might be the next big thing keep grinding

PART IV

Guys, first of all I owe you an apology. My last post was in July, and I promised to keep you updated. The truth is we’ve been working like crazy, and every time I thought about writing here, there was something more urgent on my desk. But I’m back, and there’s a lot to tell.

For the last 3 months we’ve done basically one thing: test the chatbot and the solution, fix it, test it again. Synthetic conversations are great for the start, but real fans are something else. They write at 3am, switch language halfway through, disappear for a week and come back like nothing happened, try to get content for free, try to make the model angry. Every week we went through real conversations, found where the bot was weak, and our AI team fine-tuned it. Marco and the engineers did an amazing job here again. We wanted to be sure that EVERY situation was handled.

When a fan writes, the bot first understands who it’s talking to: what kind of fan he is, what he’s bought before, how he behaves, and what he wants from this conversation. Then it adapts. A romantic fan gets the relationship he’s looking for, with no rush. A direct one gets to the point faster. A time waster doesn’t get to waste the model’s time forever.

From there it does the full job a good chatter does. It hooks the fan, builds the relationship and escalates when the moment is right, to PPV from the creator’s vault, tips or customs. It picks the content that fits where the conversation is, not just the next item in a script. If the fan wants a custom, it negotiates inside the price range the creator sets, collects a deposit and notifies the creator and the agency so they can deliver. Customs took a life to develop. It's a really variable process, and it was very hard to handle it. Not surprised other solution tend to avoid it. It also handles the hard situations: fans who don’t buy, fans who go silent, fans who try to push the model around.

During our tests we had very good results, our models kind of tripled the sales compared to the competitors I was testing but I don't to oversell. The important thing was the feedback from the early adopters that let us understood that we were doing good.

@universalofm, I remember your first comment: “no AI chat can replace a human chatter”. After these months I still respect that a lot. Our goal was never a bot that answers messages; it was a bot that sells like your best chatter, 24/7, and never gets tired or rude. I’d love for you to try it one day and tell me if we got there, or at least near there

We’re live on Fanvue, have several happy customers, both creators and agencies, using the CRM and the AI chatbot every day. Their feedback is gold for us, and every week of real usage makes the bot better.

And OF? This is the question I get the most. We’ve received a lot of requests for OF. The integration is in testing, but we’re not launching it until we’ve tested it extensively. OF is where most users are, and I’m not going to put anyone’s account or revenue at risk just to be first. Remember what I said about the competitor that was offline for 7 days at Christmas? That’s exactly what we don’t want to be. When it’s ready, you’ll read it here first.

Not everything in these months went smoothly. As you know, we’ve been growing the internal dev team, and we hired a developer who told us he was Chinese. Everything looked fine at the start, worked for more than a month. Then one day he told us his 6 months old kid was sick, needed surgery and he couldn’t work for a while. Of course we were understanding, we’re human. Then he asked for his salary to be paid in advance. Then he told us he “HAD TO” change the bank account where we paid him.

Then we finally got him on a video meeting, and it was a completely different person. Not behavioural wise, like another guy, another face, 10 years younger. After digging a bit more, we found out he wasn’t even Chinese: he was Korean, pretending to be someone else the whole time, used a fake passport, etc.

Lesson learned the hard way. Now we verify identity properly, we do video calls before and after hiring, there are no advance payments, and any change to bank details gets checked. If you’re building a remote team, especially devs you can’t meet in person, please be careful. These people are good at telling stories, and a sick kid is the kind of story that makes you lower your guard.

That’s all for Part IV. Let me know what you’d like me to cover next: the first customers, how we onboard a creator, or how we test the bot?