Discover more from House of Ventures
AI Takeover: Disruption in the Sky
The article discusses how Fetcherr, a company led by CEO Roy Cohen, is introducing disruptive technology into the airline industry. Roy, with a background in e-commerce and omnichannel retail, explains how Fetcherr's advanced generative AI technology can predict prices on Amazon's website with impressive accuracy. They decided to apply this technology to legacy industries like airlines, as it offers an edge over traditional rule-based machine learning methods.
In today’s ever-changing landscape of technological advancements, legacy or traditional industries have often been associated with a reluctance to embrace new technologies. These industries have established long-standing practices and systems that have served them well over time, making them hesitant to adopt novel methods. The airline industry is no exception,
In this episode, Omri Hurwitz is joined by Fetcherr's CEO, Roy Cohen, a trailblazer with extensive expertise in streamlining logistics and boosting global businesses. He will lead you through the exciting and disruptive intersections of Gen AI, and e-commerce in the airline industry.
Omri: Tell me a little bit about your background.
Roy: The last decade before Fetcherr, I spent mainly in omnichannel in the e-commerce industry, retail industry, e-commerce technology, fast publishing, and new kinds of ways to monetize better in the e-commerce funnel. That was my main play for the last decade. The last company that I was an employee of was STK in London, a Tier 2 mobile manufacturer. I was the director of Product and Innovation, and came back to Israel and did some e-commerce arbitrage between the biggest players. This formed the notion of opening Fetcherr, and let’s say from that point on, it took a life of its own.
Omri: You have tons of experience in the e-commerce space, which is intense and aggressive. How would you combine that with what you guys are doing now with Fetcherr?
Roy: What we’re doing now with Fetcherr is much more advanced than what we started with. I asked Uri, my partner, because he will always brag about algo trading, and I said “Okay, can you beat Amazon?” “Give me one second with Amazon and I will win the final from Google.” That’s all you need. So that’s the start of Fetcherr, if you can predict their prices on Amazon’s website. It worked when we bootstrapped and we made a machine that was able to predict Amazon’s prices for 1.5 million products, two weeks in advance with 95% accuracy.
At that point, we opened Fetcherr because we understood that we can transition capital markets algo trading technologies that are extremely fast, extremely sophisticated, the best AI, and even stronger than any e-commerce machine. So we decided to find more archaic, legacy mainframe industries, because e-commerce is a ratio of competition and go explain that you are generative AI or reinforcement learning-based AI. So we went to older industries like the airline industry. But when you bring such new tech to a legacy-based industry, the arbitrage between this tech and the tech used today with the vendors is completely different. So you see the results are quite high.
Omri: If we’re talking about technology, what sets Fetcherr apart from your competitors?
Roy: You can use ML/AI, but it will still be rule-based ML. AI, you need to do a massive amount of feature engineering. In typical ML companies or auto ML companies, you will see huge amounts of employees — 400, 600, 700, because each new customer is a project. On Fetcherr’s side, it’s very similar to ChatGPT, for example, generative AI, we have a newer network. There are no rules. There is only data that we normalize into the system and we give the system a goal. Or in generative AI or LLMs, a prompt. We have a generative market model, which is an engine that has many models. So you can tell the system, “Increase my revenue for this” and I will use e-commerce language, fair in a route in the airlines, and the system understands the market dynamics – all market dynamics, competition behavior, consumer behavior – and generates the best price in a certain time of the day that somebody will monetize. And you can put the system on autopilot and just give it rules. If it can do it, it can tell you that it can; if it does not, it will tell you what it can do instead.
Omri: What are some of the key challenges you see when you guys were trying to penetrate the market or helping the airlines adapt this new technology?
Roy: First of all, it’s new tech, so being a trailblazer, you’re always compared to great IT vendors. But our first challenge was to convince people that this tech is real. ChatGPT and all the generative AI movements helped us a lot. Luckily for us, we have great airline partners that took a chance on Fetcherr and are now reaping the rewards because they understood the tech. And being the first with tech like this in a legacy or traditional industry is huge because the rest are still using email, rule-based systems. It wasn’t an easy task, it took us 4 years. But today, the tech is validated.
Omri: What were some pushbacks or objections that you got from your first calls with prospects?
Roy: You don’t know anything about this industry, this will never happen, I don’t believe you, this is science fiction. A lot of science fiction in the first 2 years. Nobody believed this could even be doable. This problem was not solved in the 70s, but what we are now solving. But this problem was solved in the 90s in algo trading. So each industry sees things differently.
Thanks for reading House of Ventures! Subscribe for free to receive new posts and support my work.