What are We Seeing as VCs?
An investor perspective on the Generative AI craze
Please share the love with your friends and colleagues who want to stay up to date on the latest in artificial intelligence and generative apps!🙏🤖
As VC investors, we see hundreds of pitches for new companies and ideas, and invest in just a handful. In the past nine months, it would be no surprise that most of the new company ideas we’ve seen have an AI angle - particularly Generative AI. And it’s clearly not just us…in Q1 this year alone $1.7B was poured into GenAI startups, and that number will likely 5x in Q2.
In this week’s post, we wanted to share some of the common themes and pitches we are witnessing, non-trivial characteristics we look out for, and what separates “good from great” from a financial lens. The space is still nascent and nothing is definitive, but we wanted to share this post in hopes that it is helpful for future founders as they build their business and differentiate in an increasingly crowded category.
What Ideas Are We Commonly Seeing?
Early Stage (PreSeed/Seed/Series A)
At the very early stage, we are seeing plenty of “generative-native” companies being formed. These companies are natively built on top of foundation models, either as an application serving end users or the “middleware” tooling layer sitting in between the models and the applications.
Idea 1: Text-based content generation using models to create new, or enhance existing, text across emails, knowledge repositories, and other apps.
Idea 2: “Co-Pilots for X”; AI agents that sit alongside the human operator to augment their ability in writing code, drafting presentations, and other tasks. We have seen a lot of co-pilot applications for vertical-specific use cases, and a number of applications that are trying to become more “personalized” co-pilots.
Idea 3: LLM tooling for managing embeddings, and vector databases.
Summary: In January, we wrote a piece that highlighted moats in the era of AI. In order to be a differentiated early-stage GenAI startup, it’s important to have one or multiple moats. Moats can range from having unfair access to distribution, AI/ML talent, compute, data, models, or having a differentiated viewpoint on the problem space you’re solving and how to create a more delightful user experience.
Early-Growth and Growth (Series B/C+)
For the companies we are meeting at the Series B/C stage, they were typically born in the “pre LLM” world, and are now figuring out how best to leverage the power of foundation models into their existing products. We refer to these companies as “generative-enhanced” scalers that do not necessarily need to reinvent their own wheel, but ensure they don’t lose out to LLM-native startups.
Idea 1: Predictive Analytics; many scaling SaaS companies are utilizing AI to extract insights from large datasets they already have, for everything from more accurately predicting revenue growth to churn rates.
Idea 2: Personalization and Recommendation; this is one of the quickest and highest-impact ways we see growth-stage startups utilize AI. The advent of foundation models allows both B2B and B2C companies to make more robust and accurate product recommendations to existing customers.
Idea 3: “Instant auto-completion”; in virtually all growth stage companies that have a text or writing component, we are seeing LLMs being utilized for “instant auto-completion, mirroring the experience a user has with ChatGPT.
Summary: If you haven’t started playing around with ways to improve business or re-work the architecture to be more “AI-friendly”, consider dedicating a small group of your product team to build new features.
Word of caution for startups entering the space → It’s important to take stock of how much money Generative AI companies have already raised, particularly in specific sub-categories. Just take a look at this great landscape of 250+ GenAI companies mapped by Dealbook; companies building models, copywriting tools, and vector databases have already raised hundreds of millions. Of course that doesn’t mean that another innovative startup can’t be built here, but it’s important to note…
What does “good” look like from a financial perspective?
We are still in the early innings of being able to understand what “good” financial metrics look like for Intelligent Application companies, but in the SaaS world, we believe “best-in-class” growth rates mimic something like the below. Keep in mind, we are no longer in a “growth-at-all-costs” mindset, so efficiency and burn are important factors.
Time to Product Launch: One of the benefits of intelligent applications is the ability to launch a product at a faster rate than ever before. We envision that many Intelligent Application companies will launch products as they are in “beta” mode so they can start collecting user data that will be applied to create an RLHF loop. Historically, it may have taken a year after product launch to get to $1M ARR, but we may see generative AI companies getting to the $1M ARR mark even faster given how quickly customers can see ROI. Many Generative AI products also benefit from virality through the PLG / bottoms-up nature of the sale (e.g., Jasper, Lensa, Harvey, Tome, etc.)
Customer Retention: While GenAI companies may attract new customers very quickly, churn rates can be higher as well. For SaaS companies, good gross retention rates are ~85%-95% and best in class is closer to ~95%+. In terms of net retention rates, we view good as ~110% - 120%+ and we view best-in-class as ~120%+. Higher churn may result from consistently incorrect outputs by the model, 5 other competitive products springing up, etc. One important element to keep in mind with the PLG motion in the Intelligent Application use case is how easy it may be for customers to try new products or $10-$20/month for a product, but then also quickly churn off.
COGS & Gross Margins: In a previous post, we talked about how we expect many Intelligent Application companies will see new COGs around 1) the model; 2) training and finetuning; and 3) FM Ops. Since writing that post, we have heard anecdotally that costs for running queries on these LLMs and vector database storage (through companies like Pinecone) have been exuberant. In many cases, we’re hearing that customers may run queries on models until they get the output that they want, and because they are paying per license, the number of queries run meaningfully impacts costs down the line. As a result, we expect that gross margins may decline for AI-driven companies.
What separates “good” from “great”?
Just as with any other technology or sector, as VC’s we are ultimately still underwriting great teams, large markets, and a keen understanding of customer pain points. These fundamentals simply don’t change:
Customer Centric / Solving a Real Pain Point: During any new technological shift we see a number of new companies that are just trying to “ride the wave”, and are building “cool” technology but not truly solving a customer pain point. The number one question to understand is: Are you solving a “hair on fire” problem, and is generative AI a better way to help solve that problem or unnecessary?
Team - The democratization of building in this new era of LLMs has enabled many future founders to think about executing and building new products and companies. As a result, we see many founding teams that are building in spaces in which they have little domain knowledge or expertise. The question to understand is: Why is your team the best positioned to solve this problem?
Ability to Adapt Quickly and Execute - There is no doubt that this space is moving fast. It is now more important than ever for teams to be agile and pivot quickly on both product and strategy as needed. At the same time, it’s important to stick to the fundamentals and not just chase the hype. In other words: How will you react and understand when is the right time to potentially pivot your company?
Replicability - While AI can help companies get off the ground faster, that by definition means there likely be many more competitors in a category than previously. Just look at all the GenAI market maps being published and the dollars rushing into the category. Great founders and teams will recognize where there are unique holes to fill and largely avoid sub-sectors where they will quickly be lost in the confusion.
Conclusion
As VCs, we are just about as excited and optimistic as anyone else on the impact AI will have across the board. However, in the many hundreds of pitches we have sat it on in the past year, it is clear that there is plenty of hype in the category, and it is more important than ever for founders to differentiate themselves and ultimately prove product value.
A few other areas of note:
Valuations: While the broader VC market is down relative to the 2021 peak, funding and valuations for AI (and especially generative AI) are as high as ever. This is a reflection of the interest in the space by both VCs and founders, but it’s important to note that just like any other cycle (like the dot com bubble and bust), only a small % of startups will ultimately survive to an exit, and valuations one day can be cut upwards of 90%+ in subsequent years.
Gen-Native vs. Gen-Enhanced: As a Gen-Native company, what can you build that Gen-Enhanced companies cannot? As a new startup entering a category, what are you doing that is meaningfully differentiated from incumbents? BigTech companies like Microsoft, Google, and Amazon, have been moving quickly incorporating LLMs, so understanding where you can compete effectively against them is key.
Budget Constraints: As the macro environment is challenged and budgets tighten, it’s important to understand how necessary your product actually is. In the prior bull market, virtually any SaaS product could find their way to a few million of revenue. In this current environment and ongoing (though lessening) risk of recession, CIOs are looking at every line item to see what they can cut. Does incorporating AI into your product help, or ultimately not matter to them?
Funding News
Below we highlight select private funding announcements across the Intelligent Applications sector. These deals include private Intelligent Application companies who have raised in the last two weeks, are HQ’d in the U.S. or Canada, and have raised a Seed - Series E round.
New Deal Announcements - 05/26/2023 - 06/08/2023
We hope you enjoyed this edition of Aspiring for Intelligence, and we will see you again in two weeks! This is a quickly evolving category, and we welcome any and all feedback around the viewpoints and theses expressed in this newsletter (as well as what you would like us to cover in future writeups). And it goes without saying but if you are building the next great intelligent application and want to chat, drop us a line!