First to Market Won’t Protect You in the AI Age
Welcome back to Science Fair.
Last time, we talked about whale sharks and diving.
This week, we get back into the weeds of venture capital and AI.
How will AI change the way we think about investing in the business growth journey? The question and thought exercise below are inspired by conversations with Rob Olson, a venture partner with Skeleton Key and COO of PIÑATA, and Wayne Mackey, founder of Statespace. (Bios below)
For years, VCs have liked the “first to market” playbook when considering software deals because in the era of “big data”, the size and the history of the data protected companies like a moat and that moat could predict success.
Raise money, scale quickly, get to market first, and more likely than not, you will succeed.
Uber for ride-hailing and ride sharing
Blue Apron for discovering new at home recipes
Birchbox for sampling beauty products
Stitch Fix for personalized style recommendations
No matter how many competitors came up in the market after (even if they offered something slightly better) it was hard to compete.
First to market meant first to scale and first to “the data”.
Growth equaled more data, leading to better predictive models for customization, targeting, etc.
The size and history of the data was the moat.
This leg up meant more growth.
And that brought in more capital.
And more data.
It was almost impossible to catch up…unless you really f*ck up.
Now, AI is poised to create a different playbook:
Mimic successful models by interacting with them
Start with that as your base model
Add a layer of new, unique data
Leapfrog over the “first to market” companies
And with this new playbook, not only is the time to market for new startups shorter, but the capital needed for these new companies to catch up to existing models with reasonable parity will be less than the initial development cost of those first to market.
It’s reminiscent of the story of the Chinese zodiac and how a rat won the race to be the first one of the zodiac thanks to intelligence and strategy (and an ox who did it the hard way).
We already see this with ChatGPT and ByteDance. In December 2023, ChatGPT suspended ByteDance’s account because the Chinese company was using ChatGPT’s API in the development of their own proprietary LLM. Technically, ByteDance’s LLM is different, but the architecture of the model directly mimics ChatGPT’s. ByteDance theoretically can leapfrog beyond ChatGPT’s technical success thanks to these actions.
So how do you stay ahead?
Success will lie in how differentiated your data is—not the size and history of your data.
In this new landscape, your ability to evolve will drive the quality and uniqueness of your data. One way is to continuously enhance the existing data sets you have by supplementing with new demographics. Another way is to keep adding new distinct data sets.
An example: PIÑATA, a Taskflow management and optimization platform, is starting with a unique, proprietary data set that does not exist in the market. As Rob brought up, the underlying data architecture + AI-led enhancements will create an extendable and use-case agnostic opportunity. The proprietary structure of this data capture will create connective tissue between the more subjective, language-based data and the action-based data, effectively laying the foundation for a powerful Large Action Model that will allow for an AI agent-assisted workflow. In other words, PIÑATA is leveraging their unique data set to establish the foundations for a business solution that will be able to continue to evolve quickly even at scale.
So, as VCs, we should ask ourselves:
In this new AI era, do we even want to fund those “first to market”?
Historically, the race has been fixed because of the VC financing mentality.
Typically, when there is a new opportunity, VCs would flood it with a huge amount of capital driving these “first to market” companies to scale quickly. More growth meant more data. And more data meant more competitive advantage... Which meant more capital – and the ability to trounce their competitors.
But, as we already talked about, generative AI gives less capitalized but more nimble companies the ability to piggyback and leapfrog over competitors. With this possibility, we should be asking ourselves if this “big data” era framework of racing to invest in these “first to market” offerings still holds water?
Because, likely, success will rely less on how rich a company is at the beginning – either in data or in capital, but heavily on a company’s continued ability to efficiently evolve and stay differentiated.
It’s not about being “first to market”
Or how fast you scale out of the gate
Or how big your seed round is.
It’s about your ability to constantly innovate.
Rob Olson is a venture partner with Skeleton Key and COO of PINATA. Rob recently left M13 to join PINATA. Before M13, he founded the Data team at DigitalOcean and before that was the Director of Analytics & Insights at MDC Partners.
Wayne Mackey is the founder of Statespace and AimLabs, a video game analytics and training company. Before he started Statespace, Wayne got his PhD in neuroscience at NYU.
