About Your Artificial Startup
|Jean de La Rochebrochard||Jan 24, 2016|
Our dealflow is full of startups claiming that they are developing an Artificial Intelligence based solution that will be a major breakthrough in their industry.
99.9% of the time, those solutions are human-powered and even though a tech layer is apparently built and improved on top of that, it seems that most of those startups are nowhere close to crack the code of their dream product. Their fantasy is just fuelled by the current hype curve of AI that will fade away as they decide to give up and to grow a team instead of scaling a tech and a product.
If you want to be legit in that space and impress customers and investors, even early-on, you need to step-up your game by learning, staying focus on the end-game while executing the plan/solution. How do you do that ?
Parallelization: Your service will never work hand in hand with the tech because they are not moving at the same pace, not even in the same direction sometimes. Make sure you develop both your tech-layer and your service in parallel. Too few entrepreneurs in that space understand this. They end up with a bunch a good engineers powering the service with more and more tech layers that optimise the solution instead of solving the problem.
Sequencing. AI is a huge cake, not just a piece... So you need to develop one solution at a time and to make it work ! You can of course try to tackle all your NLP, Workflow, Deep Learning issues at the same time but it’s stupid unless you’re super heavily funded (which is also stupid in most cases…). It would be like Facebook in 2004 trying to be what it is today 10 years in advance. It just does not make sense. Tackle one problem at a time, make meaningful progress or you’ll regress.
Demonstrating. Stop talking about your great algorithms and the last piece of open source software that was released to solve all of your NLP issues. NLP is around the corner for about 30+ years (since Lisa basically or even before…) and I still can’t send a text without a typo mistake or ask Siri to search for an email in my inbox… We don’t ask you to build the Empire State Building in one block, we just need to see the first floor, fully operational, as a start. Show us what you can do, shoot a demo of your progress, the great complete solution will come later.
Of course this is only possible with the very best team. And not just one who can execute but one who can learn, very fast. AI and all its underlying matters are moving super fast. If your people are executing more than they are learning, you might have a problem.
Finally, because of the parallelization of the tech and the service sides of your solution, you need a strong head of product who can link the two and prioritize.
I can’t imagine a startup that will win in that space without applying those simple rules. Good luck :)