How do you come up with an idea like TaskRabbit?

For months, I'd been antsy at IBM. So I trained myself to have ideas.
I kept a notepad next to my bed to jot down thoughts in the middle of the night. If I had a pain point, I'd stop and think about it. So one night, I was out of dog food and wished someone could deliver it. Rather than going out and getting it myself, I said, "All right. Let me spend 20 minutes thinking about how I can solve this in the future."

What's the key to raising money when you're starting out?

You have to talk to everyone you know and everyone you don't. I raised my first million in seed money with a $700 plane ticket I bought for a 15-minute meeting with Tim Ferriss. We were bootstrapping the company, so it felt like a lot to spend, but you never know where those connections are going to lead.

How do you balance data and intuition when making decisions?

Thinking like an entrepreneur means breaking things down into small pieces, testing, iterating, and seeing how you can improve. Up until 2013, we had been optimizing our bidding platform, but we believed it was going to take a new model to get the metrics where we wanted them to be. Thank goodness we tested it. If we had just gone with our instincts that the new platform was better, we would have panicked when it became clear that some people were unhappy at first. But because we had the data, we stayed committed to our vision.

What would you do differently if you were starting an on-demand company today?

In 2008, mobile was just emerging. Now it's huge, and all these A.I. platforms are emerging. I'd try to get ahead of that. I'd think about where A.I. platforms are going with voice, like Amazon's Alexa and Google Home. How can you build an on-demand marketplace that plugs into these new systems? Because they are where customers are going to be spending their time.