By Vivian Cheng
A few weeks ago, Google unveiled a new version of their flagship Search product utilizing generative AI. This revamped Search experience offers AI-generated text summations of search queries and the ability to ask follow-up questions, with responses powered by Google’s proprietary LLM, PaLM 2.
In other words, it’s a lot like OpenAI’s ChatGPT chatbot — just wrapped in a Google banner.
Google is far from the only big tech firm plugging AI into their brand. Shopify, Adobe, Microsoft, and Amazon have all rolled out various AI-powered tools and features in recent weeks.
Big tech will undoubtedly play a key role in the coming AI revolution, if for no other reason than the capital they can invest in R&D. However, in their scramble to embrace AI, big tech has rolled out a slew of AI products that feel redundant, indefensible, and generally unimaginative. While the run-ups in stock prices of large tech incumbents suggests a market belief that distribution always wins, at CRV we’re more bullish on new startup AI applications.
We believe the most delightful innovations in AI are coming from the bottom up, not the top down, as companies like Midjourney, Runway, and Stable Diffusion have captured the public imagination. Open source horizontal models like ChatGPT will form the foundation of the AI boom, but many more venture-scale businesses will be built on the application layer.
If this is indeed the future, founders must be judicious about whether their idea is defensible, and therefore venture-backable. Here are some recommendations for such founders, based on what I’m currently seeing in the market:
Ship Quickly and Prioritize Ruthlessly
Speed is arguably the one definitive advantage all startups have against incumbents. Midjourney and DALL-E were early to shipping great horizontal platforms, but Midjourney has come out on top thanks to quick product execution. This is why one of the biggest recent AI product releases, Adobe’s Firefly Suite, barely made a dent in Midjourney’s traffic. As we speak, the 12-person team at Midjourney just dropped Midjourney 5.2, with impressive new features like “zoom out”, outpainting, and panning.
Build Novel, Best in Class Interfaces
Get creative and think about what new product paradigms and interfaces are now possible thanks to AI. This will be much more defensible than a product/tool that an existing software company can easily add onto their product suite.
Two examples: Inworld (a company CRV first partnered with in 2021, when we co-led their seed round), Character.ai, and Synthesia (AI avatar video generation) are examples of new product paradigms only possible due to AI. Additionally, ChatGPT’s success is a case study in using advances in NLP to drive a shift change in UI/UX. While the underlying technology, GPT-3, already existed, the conversational interface OpenAI built on top made GPT-3 accessible to a mass market.
Build Community
Successful products in the AI space can start out looking like consumer companies. The largest Discord communities right now — Midjourney (#1 at 15M members), Open AI, Blue Willow, and Leonardo.ai — are all consumer AI companies today.More users means more proprietary data to ship better product. Midjourney is a prime example of how amassing a large audience makes you defensible when an incumbent tries to step on your turf (i.e., Adobe).
Of course, this presents a chicken and egg problem. By definition, startups don’t have the user base necessary to inform defensible product development. But startups can capitalize on the excitement around trying new AI applications right now and a first-mover advantage.
Once a business has found product-market fit, there are a variety of techniques to foster community. Here are a few examples I’ve seen work well with founders in CRV’s AI portfolio including Inworld and Vercel:
- Grow mindshare by developing an online personality towards issues relevant to the AI community (development, regulation, etc.).
- Create dedicated spaces for users of your product to engage with the brand and each other (e.g., Discord, Slack, Reddit).
- Offer perks to existing community members (e.g., gated access to beta users through Discord to grow the community, referral codes) to foster customer evangelists and incentivize non-members to join.
- Appoint power users as moderators in branded community spaces.
- Gamify the product experience with a leaderboard or rewards program.
- Create a product with shareability and virality in mind.
Go Vertical
Startups can win initially by moving quickly, but long-term moats are built by going where big tech and incumbents won’t. In a world where anyone can access LLMs, verticalized solutions based on a deep understanding of a target persona is one of the best defensible ways to build in AI.
Consider EvenUp, which recently raised a Series B led by Bessemer Venture Partners at a rumored $325M valuation. Where big tech went broad with their applications, EvenUp went narrow by developing an AI-powered tool that helps personal injury lawyers collect information and file claims. EvenUp is a great example of how startups leveraging proprietary, vertical specific data and sticky workflows can build long-lasting moats.
I’m particularly excited about vertical applications where buyers have the budgets, and the outputs are more subjective vs. objective (i.e., image and text asset creation use cases in marketing, architecture, game studios, etc). Moats are cemented by building with collaboration, integrations, permissioning, and workflow top of mind. While we think industries like legal and medicine are also ripe for disruption, entrants will need a strong solution around hallucinations, as accuracy is crucial in these industries.
Market size
While we believe vertical applications will drive widespread AI adoption, there’s such a thing as niching down too much. The opportunity still has to be large enough to drive venture-scale returns.
To determine if the market size is suitable for a point solution, we recommend a top-down and bottom-up analysis. A top-down analysis entails determining the largest possible addressable market (TAM) based on information currently available and key assumptions about your business. A bottom-up analysis estimates how much you can scale the key “units” of your business (product, price, and customers). The size of your market typically lies somewhere between the results of your top-down and bottom-up analysis.
When evaluating the market, it’s also important to understand the buying power of your audience and their actual needs. For example, say you’re aiming to develop an AI co-pilot for doctors. The medical profession is notoriously difficult to sell into without a ton of budget and a tolerance for red tape. While someone will eventually break through, it’ll take a considerably longer amount of time than selling to, say, marketers who are already seeing an ROI from AI tools.
Conclusion
The only truly defensible items in the application layer are workflow, product, community, and rapid iteration. As such, founders hoping to thrive in the coming AI boom will need to ship quickly, create a differentiated approach, and focus on building community around their point solution — then use the data collected from their community to stay differentiated. Many will need to further niche down (while still pursuing an economically viable market),
Startups have the flexibility to be more creative and nimble than big tech, and that is a key advantage in the early days of this AI revolution.
If you’re currently building in the AI space, I’d love to talk. It’s the perfect time to build new applications in AI with bottoms-up, product-led growth. To get in touch, find me on Twitter or reach out by email at vivian@crv.com.
