User-Centric Growth: From PMF to 720,000 Users

Insights from David Park, CEO of Jenni.ai

Happy Saturday to the 7,787 builders, marketers and unicorns reading this newsletter.

Today we're going to dive into 3 strategy insights from David Park, founder of Jenni AI.

Jenni.ai is an A.I. writing assistant that's amassed 720,000 users and $94K MRR from organic social.

David and I had an amazing conversation that ranged from their origin story to long-term positioning in the A.I. space.

I've cut down the details into 3 key ideas that I think you'll find really useful:

  • How to find product-market fit by reducing complexity

  • Using organic social for growth

  • Strategic positioning among a flood of A.I. competitors

Time to read: 7 minutes

Difficulty level: HARD

If you're new here, join the unicorn tribe for $0 with one click. 🦄

"We really only started getting traction when we decided to cut all the bullshit and said "'what do users really like about our product?'"

- David Park, CEO Jenni.ai

🦄 Growing Your Engineering Team? 🔥

If you need 20+ hrs/week of developer time, Lemon.io has vetted devs for you.

You'll get a developer match within 48 hours and a free quote.

Sourced from 40 countries, less expensive than lower quality alternatives. Try Lemon.io!

Now on to the main event:

What is Jenni A.I.?

Jenni.ai is an A.I. writing assistant tailored to the student and academic market. It provides auto-complete recommendations as you type and has a built-in plagiarism checker and citations functionality.

Jenni.ai focuses on “human in the loop” and augments humans rather than trying to generate the entire piece of text for you. Unlike ChatGPT, Jenni AI queries the web in real time and allows you to add citations for the underlying sources it uses as it writes.

They offer a freemium model that gives you 200 words per day, free. Try it out.

Here are summarized notes from my conversation with David on the evolution of Jenni.ai as it evolved to its present state.

Finding Product-Market Fit By Removing Complexity

Three or four years ago we were really interested in natural language processing. This was before GPT-3 was out. There was only GPT-2. But even then, I couldn't believe that an AI could generate a sentence. Sure, the quality was terrible, but the idea that a machine outputted text was mind blowing to me. I thought - if we just get this to a better level, then this could really change the world. So even when there was only GPT-2, we were trying to train our model. We always had this vision of AI-assisted writing.

Empowered by the low-quality GPT-2, we created a content marketing agency that leveraged A.I. to deliver content to marketing agencies. We delivered over a million words of content. Unfortunately, very little of it used A.I. We were so enamored by the possibilities. that even though so little of the AI outputs were usable, we just had to incorporate it into the business.

When GPT-3 came out, that's when Henry (my cofounder) and I were like 'holy shit', this is a once in a lifetime opportunity. One of my favourite quotes is: “Opportunities come infrequently. When it rains gold, put out the bucket, not the thimble.”

We were just too hyped about the technology.

We started doing complex things.

With our first product iteration you would put in a title and get an immediate outline of a complete article. As you changed anything within the top of the article, the remainder of the article would update in real-time.

To us, this idea sounded very appealing/sexy. It was like a living, breathing document. And I was pitching it like that. Some really advanced users could figure it out.

But most people were like 'What the fuck? I just changed a word and suddenly everything changed in my document, I don't understand what's happening.'

We really only started getting traction when we decided to cut all the bullshit and said 'what do users really like about our product?' 

And the really sad part was that there wasn't much about it. Basically, when they get a generation that completes their thought and saves them a little time.

So we literally ripped out most of the functionality.

We had a keyword finder.

We had multiple panels with different writing modes.

We had the living breathing document.

We cut everything out.

For weeks, we'd cut something out and conversion rate wouldn't change. People didn’t care about most of our features.

We removed so much complexity, until it just became a document with auto-complete. And that was when users were like 'I love this.'

At some level, it's a little insulting to our ego. I was trying to be on my Steve Jobs' ‘visionary’ shit and our users hated it. And then I do something that essentially already exists (similar to Gmail auto-complete) and people started spreading Jenni like wildfire. That was quite a humbling experience.

We ripped complexity out and delivered on 'Can A.I. assist your writing? Can it auto-complete what you already want to write.' Once we kept going in that direction, then things started taking off.

Since then we only listen to our users.

FREE: Get Your Startup in Front of 7,787 People

Have you achieved a milestone in your own business recently?

We want to use this community to support our readers who are building cool stuff.

The first 3 people to respond to this email with a noteworthy and credible milestone will get a free mention of their startup in next week's newsletter.

Example:

"My startup, awesomedomain.com just hit $5K MRR through paid ads."

So reply now and tell me what you've done!

Check back next week to see who made the cut.

Now back to Jenni.ai...

Getting 720,000 Users From Organic Social

When we pivoted from a content marketing agency to SaaS our MRR dropped from $7K to $2K.

We just could not escape $2K MRR for 2 years.

I'd gone viral on social before Jenni and I knew we could crack it.

Our product was at a place where even if we did go viral, they weren't going to convert. And even if they did convert, our product was confusing.

We really had to figure something out before we focused our effort on going viral.

Our conversion rate was almost essentially 0%, and then went from 1%, to 2%, to 3% and then 4%. And that was a number we could work with.

I didn't want to waste time engineering virality and then not getting conversions. If we could make a great product, it would help virality.

Our primary growth channels have been:

  • Social media (6-10M organic impressions just from our own accounts each month)

  • Word of mouth from people who enjoy Jenni and recommend to their friends

When you have a product that's visually appealing, like an A.I. writing for you, it's killer for short form videos to go viral.

The Big Break

Jenni appeared on the most viral Twitter thread in history:

That was our first big break.

It literally took our MRR from $2K to $10K in one month.

I didn't even talk to Zain. I'm not sure how he found us.

We'd only made a few TikToks, we'd sponsored a few influencers, and it was one of the luckiest breaks ever.

TikTok Conversion Rates

The fact that 6 million people view your video, you're bound to get a few users.

If you get like 60 new users from millions of views, that might seem really low. But if you only spent 1-2 hours putting that TikTok together then it all evens out.

Organic just doesn't convert that well. It's really a volume game.

One thing about organic is that you can't control where you go viral.

The U.S. used to be our biggest market. Over the last few months we've been spreading like wildfire. And that's been dragging down our overall conversion rate.

Positioning Jenni.ai For the Student Ecosystem

We had a fully functional B2B product and we ripped it all out.

Scaled it down to a little kernel. And then scaled it back up to a completely different B2C product.

And that's when students started to love us.

You'd think that students would want a complete essay. That they'd want the work done for them.

Yet our conversion rates kept getting higher the more we encouraged humans in the loop: 

  • The more we lagged it so you don't get the next auto-complete immediately, the higher the conversion rate went

  • We lagged it so even when you get the next auto-complete, we force you to read it

  • You can't accept the suggestion right away

When we force you to read it, students are much more likely to convert, because students have to actually interface with the A.I. Maybe they don't like the suggestion, so they re-roll it or they just type it themselves. These are the types of decision where we're putting more power into the human's hands. Overall that's just been a good direction for us.

A year ago, students hadn’t seen this tech before. They didn’t know that an A.I. could write this well.

One of the biggest asks we got was for a plagiarism checker. Students couldn't believe we didn't just rip this from the internet somewhere. So we partnered up with CopyLeaks - they service a lot of US universities. Nowadays, the market has evolved so much that plagiarism checking isn’t as highly requested as “AI detection” so they don’t get in trouble with their schools for writing with an AI. Another feature that students asked us for - 'I'm going to turn this in. I need to know if this is valid or not. Can I see where the citations come from?'

So we built out this whole system - where you can highlight any text in your document, and you can see what sources the A.I. consulted to fact check it.

Now you're not just writing faster with A.I., you're learning while you write.

Example citations within Jenni UI:

Quick Product Iteration From Student Feedback

We've been adding new features constantly that students are asking for.

Students want to sound more academic or they need to sound more like their professor, so we made a feature that allows you to change the tone.

Students like the grammatical underlines on Microsoft Word, so we went and partnered with Grammarly and now have the Grammary plugin within our product.

A.I. writing can be a hindrance. It could make you lukewarm as a writer if you're not careful.

If you keep the writer in the driver seat, and you're careful, and you only augment their knowledge, then you can end up with something really beautiful.

The weird thing is that higher auto-complete rate doesn't equal higher conversion rate. There's a very weak correlation rate. We think it's because if you're not that invested into Jenni, you probably just want to use up your quota and see what comes out.

But if you're actually writing something of high quality, you're typing it, waiting, seeing what comes out, re-rolling it. Then you accept it, you edit it a little and then you type the next sentence. And now you're in flow.

When Jenni works, you're not auto-completing as much. It's really unlocking flow. Less than 30% of an exported document is A.I. generated.

Our big killer feature is that we generate words for you. But when they export the document, less than 30% is A.I. generated. Of any stat, that's what I'm most proud of.

The biggest temptation for anyone is 'Well that's kind of what I wanted to say.' And you just let the AI write and drive your document. The result? You end up with a lukewarm version of what you wanted to say. And there's nothing more dangerous to the collective intelligence than the dumbing down of all content on the internet.

Users are able to write faster, for sure. But clearly the majority of all text was written by a human.

Our product is killer among Masters and PhD students.

The people who are writing high volume of things that are going to be published in research journals are using Jenni.

Grads have a 4-7x conversion rate than undergrads.

The Competitive Landscape in A.I. Writing Tools

I'm not scared of these A.I. tools popping up on Product Hunt.

My focus has always been placed squarely on when the incumbents fight back:

When MS Word and Google Docs gets A.I. writing capabilities.

That's when you'll see a culling of 90%+ of the market. All these GPT-3 wrappers that are features and not products.

To survive, we have to really zero in on the student market and be the platform where students come and get the most value.

We can't just be an essay help tool. We need to be a student augmenting tool.

Our services have to become so compelling that students choose not to write their essays in Google Docs. And we're already seeing people do that.

How can we make it end to end? A.I. assisted notetaking, note cards, make the lives easier for students.

The branding and positioning and ultimately the value we have to give is that 'I become a better student with Jenni'. And notice that I'm not saying that we have to have the best generations. When Google Docs and MS Word have A.I. writing, they'll be as good.

But we have plagiarism checking, citations, A.I. detection features - all these things that Microsoft and Google aren't going to have, because they're a broad writing tool.

A lot of these A.I. writing tools grew fast just by wrapping GPT3. And unfortunately, they'll probably fall fast as well.

Thanks David for sharing Jenni.ai's evolution with us. If you'd like to augment your writing with A.I., check out Jenni.

Presented by:

A Message From AdQuick

Out-of-Home Advertising, Made Easy & Measurable

AdQuick seamlessly connects advertisers to out-of-home (OOH) media owners anywhere in the U.S. and abroad, and makes it easier than ever to plan, buy, and measure every kind of outdoor advertising –– from static billboards to hand-painted murals. If you’ve run online ads you’re used to the speed of deployment and accountability provided via analytics. AdQuick provides the same experience digital marketers are used to to the OOH space.

With AdQuick, advertisers gain immediate access to over 1,400 premium OOH media owners, can execute OOH buys 10x faster, launch OOH campaigns in as little as 48 hours, and unlock unprecedented measurability.

Learn more and get started at AdQuick.com

A final question for you...

This is by far the longest issue of Unicorn Growth Strategies so far. But I had to leave out a ton of good content from my conversation with David.

I need you to weigh in on what you'd prefer each week:

What's your preferred content format?

Login or Subscribe to participate in polls.

✨ That's it for today.

My conversation with David inspired me that the long-term winners in the A.I. space will be those who successfully augment human creativity, rather than flood culture with low quality crap.

I hope you enjoyed the read.

-Brian 🦄

Btw - Jenni's story is a great example of fast moving water. Worth a read if you haven't seen it yet.

Recommended Reading (if you subscribed recently):