The best AI startup ideas won't be selling AI
How growth hacking with AI will be the future
It’s hard not see the hype in AI after seeing Jasper.ai raised 125 million for a 1.5 billion dollar valuation last week.
Their overall growth in copywriting AI is staggering. And it got me thinking - is it really that easy to just build a UI on top of GPT-3 to build a billion dollar business?
For everything that I wrote about AI business defensibility - it seems like Jasper has bucked the trend by making themself the standout out of the plethora of copywriting startups. Copy.ai is rumored to do around $10 million in annual revenue while Jasper.ai has done much more. Also raising $125 million will also surely prompt the attention of other copy-cats to enter and get a slice of the same market.
And from their press release it sounds like they’re set to try to differentiate themselves through the classic feedback loop model I talked about before.
“To build a market leading company in generative AI we need the appropriate infrastructure — that’s what we’ll raise [new] capital for,” Rogenmoser told TechCrunch in an email interview. “We want to build a world-class business, [and] to do that we need capital and highly strategic partners.”
….
Rogenmoser claims that Jasper’s language models — trained on 10% of the web and fine-tuned for “customer specificity” — set it apart. Among other apps and services, they power Jasper’s browser extension for Chrome that delivers contextual content recommendations across platforms including Google Docs, Gmail, Notion and Hubspot.
“The folks that will win at generative AI will be the ones that have the best feedback loops,” he said. “We’re committed to building the best feedback to AI loop.”
I think this is more marketing hearsay than actual product. They have their work cut out from them in building this “customer specificity”. But I can understand why people invested for a different reason.
Their feedback loop is not so much building the AI - but instead I believe through using AI to build the growth engine internally.
Jasper’s Historical Growth
By tracking Jasper’s growth across Ahrefs for SEO - we can see that in a span of a year or two, they managed to increase their organic traffic to 150K which is a pretty impressive feat. But especially also since they only raised a 6 to 8 million dollar seed round beforehand.
To be fair - they rebranded from Jarvis.ai - so it looks like they started their growth over a year ago instead of 6 months ago. But still impressive.
By comparison - this is my company, Interview Query’s organic traffic growth without using AI tools (and also without seed funding btw).
Through using their own AI copywriting tools, Jasper was able to fuel their own growth by exploiting various capabilities in Google’s SEO algorithm (likely amongst paid ads as well through copywriting but that’s a little harder to track).
I use the word “exploited” in a loose term. But they found ways to increase the efficiency of their own content marketers, to build a growth engine that could increase their SEO presence faster and cheaper than any other company selling an AI content marketing platform.
Specifically here, I would note their feedback loop then is not really in their middleware model layer on top of GPT-3 so much as in refining their own growth engine internally to grow their business.
A growth engine has historically been built through a mix of efficient operations and strategy. You test different growth strategies, double down on the ones that are good, and execute on it faster than your competitors.
And while Jasper sells growth tools to other businesses, they have the advantage of dog-fooding their own product to figure out which middle layer tools and human + AI interactions work best towards pushing growth.
So previously when I was looking at my own mental model for businesses - I saw something like this:
OpenAI → Middle Layer Application Model → Sell to Customers.
Jasper instead I believe is doing this:
OpenAI → AI Growth Engine → Find Customers.
The AI growth engine is what investors are betting on. It makes sense that investors see Jasper not as a complete differentiator in the market, but rather a company that can use AI to grow faster than their competitors. If this is a winner take all market, which I’m not sure it is, then investors believe Jasper can build their growth engine faster than anyone else.
Growth Hacking AI Businesses
Like every growth engine, the biggest risks comes at the expense of the platforms they’re serving to grow. Once Jasper themselves can’t find ways to increase their own growth through Google, then their product becomes more or less useless. And this isn’t even mentioning the problems with market saturation as other tools come into popularity and take market share.
For example, when I joined Jobr, a Tinder for jobs type company, one of the reasons we became so so successful was because the company had started in the same period that saw low customer acquisition costs from the emerging Facebook mobile ad era. Once Instagram mobile ads were launched a few months after I started, we used it to spearhead our growth into a timely acquisition before Instagram ads were mainstream.
So like any edge found when growth hacking, there’s almost always a saturation point to which it stops working. And the question with Jasper and other AI tools if if they can be continuously tinkered with to find new edges to exploit and grow businesses to gain market share.
With this fundamental assumption, there’s an argument that the next generation of AI applications may not actually need to be selling AI at all. Most people, including myself, are always first fascinated by the fact that AI tools could be sold like SAAS tools with new businesses for therapy and coaching, replacing interior designers, or creating in-game characters with side quests and more.
But rather, Jasper has proven that the argument for using AI growth engines for normal businesses with product market fit is likely a better strategy
By utilizing AI under the hood as a growth engine, you could build a business in any growing market and with enough market share for new entrants. Your advantage would solely be in your ability to grow faster than competitors and to exploit AI to continuously find new avenues for growth. In a changing landscape where your growth edge gets nullified rather quickly in startups, AI’s best use case is being able to find new growth trajectories faster than your competitors.
For example, I’ve been seeing a lot of dating chatbots enter the space because the technology is now here to make creating new pick up lines easier. Yet the unit economics of dating chatbot businesses don’t look very good given high churn and low consumer SAAS pricing.*
Instead the new best ideas are likely to find a normal business with strong retention and high CAC and growth hack it with AI. If you can exploit the AI growth channel before it gets saturated, then you might be able to use it to get to some X% of the market as fast as possible.
I see a couple new ideas for businesses here. If you can bear the stereotypical gold rush metaphor:
You have the companies like Jasper and Copy.Ai selling shovels in the gold rush. Basically building growth and revenue facing AI tools.
You have the businesses leverage these tools as the companies mining the gold. So this could be a marketing agency that uses these tools to execute on their clients strategies better. Or just any company that wants to grow faster.
And lastly you have creators, teachers, and agencies that could sell courses on how to hire diggers and get them the latest AI shovels to mine gold a little faster.
All in all - it really is an AI gold rush - and we’ll see who profits from what in the near term!
[1] For example → if we assume $100 CAC and then $10/month product with 20% churn. The unit economics don’t exactly work out. We can run the math on this.
10 + 8 + 6.4 + etc… < $100 total customer lifetime value.