Navigating Hype: Insights from the Self-Driving Industry's Triumphs and Failures for Generative AI Entrepreneurs
In 2016, with the launch of Uber's driverless division, Zoox and Argo raised over a staggering $1B in capital each. Cruise was acquired by GM for an equally impressive $1B. Investors, both strategic and venture, hailed self-driving technology as the impending mega-disruption. But, the clock fast-forwards to 2022, and the self-driving industry saw a $40B loss in market value. Startups either ran out of capital or pivoted their business models.
As Generative AI rides a similar hype wave, exemplified by the successes of OpenAI and Chat GPT, and recent mammoth $1B investments in Inflection, Anthropic. MosaicML's was recently acquisition by Databricks for $1.2B. While there are similarities and differences between the two fields, it’s imperative to learn from the missteps of the self-driving industry.
Lesson #1: Prioritize Revenue Generation Early
In the self-driving industry's formative years, a key debate was whether to prioritize fully autonomous mobility (sans safety driver) or highway autopilot systems (with a safety driver).
While full autonomy promised grand visions of trillion-dollar companies, thanks to the elimination of 80% of on-demand transportation costs linked to drivers, it presented a technical challenge ten times harder. In contrast, the autopilot approach offered a more technically feasible path with a safety net of a human driver but with a smaller market.
Most startups chose to chase the autonomy grail, but soon, a mere handful of players remained globally as capital dried up and investors shifted their focus to cash flows. These companies hardly had any recurring revenue.
In sharp contrast, Mobileye, focusing on the gradual approach of highway autopilot systems, saw steady revenue growth. By 2022, they held 70% of the global market share for advanced driver-assistance systems and raked in $1.8B in revenue. Their robust cash flow even funded their foray into developing fully driverless cars.
Generative AI founders can draw a valuable lesson from Mobileye’s success: Prioritize immediate revenue-generating opportunities and hasten the journey to positive cash flow.
Why?
In industries susceptible to hype, capital often dries up unexpectedly, and venture funding ceases. Investors then pivot from betting on vision and team to scrutinizing business metrics and unit economics. If your business is revenue-positive when this happens, you can switch to offense, even contemplating acquiring competitors. However, if you've neglected revenue generation, convincing customers to pay for your products becomes ten times harder.
How to identify revenue opportunities?
Seek out your customers' burning pain points and build solutions. Quantify the business value you bring (like increased productivity or decreased costs). Paid usage is a telling metric for startups—it denotes the value your product provides and indicates its worth to your customers.
Lesson #2: The Last 5% in Product Development: Significant Yet Challenging
The last 5% of technical issues often present the greatest hurdle for self-driving technology, such as differentiating a balloon from a brick on a highway at 60 mph. Yet, addressing these challenges is crucial.
How does this relate to Generative AI?
Building generative AI tools—from agents for API integration to meeting summarizers or workflow automation tools—requires your customers to rely on your product as a part of their daily workflow. The last 5% of product work and features to delight your customers is incredibly important. As your customer scales, you’ll likely have issues with scalability, security, and reliability. There are no shortcuts, and this takes a culture of customer obsession.
One thing I learned from my co-founder Chris, is that the best engineers can both understand research papers and translate them into real-life technology and product. Finding folks that are willing to put in the hard work to fully solve the customer issues and feel enough ownership to fix problems when the services break, are gems. Many of the pioneers in generative AI built technologies from research labs and academic institutions. Pairing them with folks that have brought successful products to hundreds of millions of users will maximize your chances of building a generational business.
In the early days of a startup, I believe that everyone at the company should be talking to customers. If it’s an enterprise business, set up a joint Slack channel. Do weekly check-ins, and monthly product roadmap sessions. All of these things help you build trust with your customer.
In conclusion, if you are running a generative AI startup today, (1) prioritize revenue generation even before your investors ask for it and (ii) obsess over your customers until they incorporate your product into their everyday workflows.
My hope is that we can take the lessons from the self driving failure, and use them to build the next generation of trillion dollar companies!
Thanks for sharing your learnings from the self-driving wave. They seem to apply well to this current wave of generative AI.