January 11, 2017

innovators: interview with David Barrett of Expensify on Launching an AI Product

launching-an-ai-product

Earlier this year Clearhead’s favorite expense report app launched a major update focused on AI. Being in the business of helping our clients optimize digital products, we had to learn more. So we sat down with Expensify’s CEO, David Barrett, to learn what this new update was all about. In short, it’s called Concierge, and is an artificially intelligent virtual assistant built right into Expensify. That means Concierge knows everything Expensify knows — which, given the billions of dollars in expenses they process, is a lot — and is eager to use that knowledge to make your life easier.

In fact, Barrett went on to tell us that it’s possible if you are an Expensify customer you’ve already spoken to Concierge without realizing it. Why? Because her first job at Expensify was customer support.

This AI product is also the brains behind their “Price to Beat” feature, as she not only knows every expense ever reported by anybody, she also knows the real time pricing of every flight, hotel, and car in the world. This means Concierge will discreetly notify you if you’re getting the best value for your travel dollar — something that matters even more if your company has enabled “Expensify Rewards”, as you keep $0.50 of every dollar you save your company.

We sat down with the Founder and CEO, David Barrett, to talk about how he developed this product and the test and learn methodology that went into it. Here’s what he shared with us.

1) When testing and building a product, what’s the biggest mistake people make?

The most important thing (and the biggest mistake people make) is taking the time to truly think through and understand the problem you are trying to solve. Too often, people get enamored by a particular technology and invent some kind of problem for it to solve — without really confirming real people experience that problem, or that it’s the biggest problem they are experiencing. This produces a lot of hard work and cool features that gain no traction in the real world because nobody cares.

2) When launching a product, what’s the key to a solid go-to-market strategy?

I’d say the keys are:

  • Identify a common, frequent problem that is sufficiently severe that people would pay to have it solved, and that you are also experiencing yourself. This way at the very least you know you have one person who will like what you build: you.
  • Design the simplest possible solution to that problem
  • Split that solution up into at least three parts, with the first part being the minimum you can even conceive of showing to anyone — call that the alpha, and test it yourself. Get it working to a point where you are personally happy with the feature.
  • The second part is the minimum you can imagine showing anybody else — call that the beta. Show that to whoever you trust most who is actually experiencing that problem. Don’t show it to anybody who doesn’t have that problem because their feedback will be distracting.
  • The third part is what you are ready to show to the world — it is the most minimally scoped, basic thing you can imagine anyone else really using.
  • Promote that to all the people you identified in (1). (If you didn’t identify anybody in (1), then start over.)
  • Expect to be wholly disappointed with the reaction you receive, but keep at it! Don’t do it if you aren’t so excited about it that you are committed to seeing it through.

3) What are the top three to four things that people launching/building product could learn from what you just did with Concierge?

I think the main things we did right were:
1. Focused it on real problems. It’s really sexy to create a general AI to have long conversations with, but nobody really wants that. We focused it on answering tech support questions, and grew from there.
2. Launched it very slow. Concierge was in production and helping users for nearly a year before we talked about it.
3. Kept at it. It’s not done; not even close. We have years of active development to go. We’re in no hurry to make the first AI assistant. We are in a hurry to make the best.

4) What would be the top three things you want entrepreneurs and intrapreneuers to know before launching their product?

  1. Focus on the customer problem, not the technical solution.
  2. Take it very slow.
  3. Don’t give up.


Want to know when we share our latest content? Subscribe to our blog.