We’re living in an era (once again), in which startups are taking on some really gnarly technical challenges. These companies are building self-driving cars, high-speed transportation systems, AI autonomous agents, and mining 99% of the genome that hasn’t really been understood yet in terms of how it affects the human body and disease.
Many of these ideas aren’t exactly new. We’ve been exposed to the self-driving car (Minority Report), AI that can perform a task start to finish (Her), and robots that can move like humans (Ex Machina). What’s new is that it’s now possible to implement some of our wilder technological dreams. As a result, today’s startup landscape includes everything from products that came out of a hackathon and launch in a month to major ten year-long moon-shots and everything in between.
That in between feels kind of new again because we’ve spent the past decade focused on “lean startups.” And during that time, a whole industry developed around teaching entrepreneurs how to be lean. The core tenets are familiar: get an MVP (minimum viable product) up and out the door as fast as possible (usually within weeks), get people to use it, iterate, get to v1.1, start amassing customers, learn and iterate from there. At this point, you might raise a few hundred thousand dollars and hire someone who’s not building product. If you’re lucky, you’ll do a little bit of marketing (aka growth hacking) and watch that growth curve go up and to the right. Only then do such companies usually consider monetizing their customer base, which means identifying and building the additional features customers might pay for.
The Y Combinators of the world have further popularized this particular way of creating a product and building a company around it. And the process has worked well for the likes of AirBnB, DropBox, and Stripe. It works particularly well when you can identify some inefficiency or pain in the market (unused spare rooms, file sharing systems that don’t work across devices, inefficient online payments) and develop software to eliminate it.
On the other extreme are companies like SpaceX, Tesla or 23andMe. These moon-shots redefine what we think is even possible. No one would ever suggest that Elon Musk abide by the lean methodology. There is no notion of a buggy but usable v1.0 electric car; at the failure rate of a typical MVP, such a vehicle would harm thousands of people. No, these are hardcore, capital intensive technologies that resist current startup trends because of the audacity of their vision.
But there is a middle path between these two extremes. That’s exactly where a bunch of AI companies dwell. There’s us of course. We make an AI personal assistant who schedules meetings for you. And there are many others. Clarifai is building visual recognition software, making computers see the world. Banjo lets you understand what’s going on anywhere in the world in real time. The more general purpose Sentient provides an AI platform that can be used to solve problems in a variety of industries. What we all share is that data and AI in particular is at the core of our products.
There’s nothing wrong with lean but when you’re pushing the boundaries of science, it fails miserably. Building an AI autonomous agent, for example, is not for the faint of heart; it has required about three years of R&D. (We’ve written about it here and here.) And research is just another word for really smart people + time which directly translates into money.
This fact has some very tangible consequences. For one, you need to raise much more money up front to hope to build the most ambitious AI products. While a lean startup might raise a few hundred thousand after launching an MVP, we raised ~$2M just to find out whether meeting scheduling is a tractable problem—that is, could we ever hope to train a machine to take over this task, start to finish?
All that fundraising means that you have had to align many smart people (VCs) to bet on you and your vision. And this forces a certain discipline. There is no vague notion of finding your customers sometime in the future. You need to truly understand the pain and the customers who have it early in the journey.
This gets to the another odd dimension of these anti-lean ventures, which have what lean startups sometimes lack, customers willing to pay for them. This fact alone upends the usual freemium model! Don’t get me wrong, I’m all for lean. It’s just that building an anti-lean company allows for ambitious long term product development, which is how you build a truly intelligent agent in 2016 (or any other ambitious AI startup). It took us almost three years, more than 80 people and $34 M in raised capital to get our product to market. Paul Graham and Eric Reiss would shudder at the thought. But we launched to a pool of eager customers willing to pay, and we think the sky’s the limit from here!
In conclusion. Not everything can be solved in 3 months and you should allow yourself to be super ambitious!
Cheers //d. 🙂
This post originally appeared on LinkedIn Pulse, here.
Want to hire Amy + Andrew? Start your 14-day free trial HERE
Like What You’re Reading?
Sign up for our weekly newsletter
to stay up-to-date on the latest in AI.