- We're on the cusp of something big. Artificial intelligence is entering every place where there's a decision, often one done on autopilot, and turning it into the optimal one done with an AI co-pilot. Today I'm sitting down with Pratap Ranade, CEO and co-founder of Arena, and he's gonna show us one way all pricing will go from the domain of a person taking wild guesses in a spreadsheet or using a clipboard into one that is driven by instant optimization and having an always-on AI means billions of dollars in extra revenue for some of the biggest companies in the world. Let's get started. (upbeat music) (mellow music) Most people seem to think about business processes, you know like pricing, the way they wanna make a home-baked chicken at home, like in that old infomercial. - He had his rotisserie oven and he said, Set it and forget it. - Set it and forget it.

- But business is not a rotisserie chicken and shouldn't be run like a mechanical process where there's no room for decisions. Back in the '80s, Toyota came along and showed us what you could do by making a vastly better car, pushing decision-making into the people who actually do the work. They called this at the time the Toyota production system. In 1986, Toyota became the first import automaker to sell more than one million cars in the US in a single year. Kaizen or continuous improvement made Toyota one of the world's highest quality manufacturers. Today we're hanging out with Pratap Ranade, CEO and co-founder of Arena AI, a company funded by Initialized, and he's also a YC alum I've known for almost a decade. Pratap's company has found a way to make massive improvements in business. Think double-digit percentage growth, not just a percent here and there.

These changes are so significant. I think they have the ability to change the trajectory of entire industries. - We focus a lot on the few decisions that we do make or we can make at a high level and try and get the perfect answer, the perfect decision. And I think your Toyota point is spot on, which is instead, what if you could have autonomy and agency to make millions of little micro decisions, precision adjustments? You don't know what's happening, but you can adjust when the assembly line stops? And I think that hasn't made its way out to a lot of other industries. (mellow music) - So how does it work? Here's Pratap showing what his software actually does, enabling micro decisions and precise adjustments to make those decisions happen continuously. - With pricing, for example, what you might do is say, I'm gonna run a 50% off on Black Friday. That's a consumer example, as you said.

So that's basically the fixed point for everybody. But instead, Gary right now might actually not care about that. Maybe it's actually he's considering subscription and a discount on a subscription purchase is the right thing for Gary. Could be an entirely different setup where you could say, Great, I wanna have a bundle offer. I wanna have loyalty points. And so what happens is you have this incredibly wide action space of possible decisions, and because of our own human cognitive capacity, we only make a subset of them. - Building great software is about empowering humans to be cyborgs instead of brainless robots in a black box. Pricing is just one of the most profitable examples of a decision that is normally poorly optimized and made by humans. Pratap and I met at YC as he was going through the program and I was a partner there. He sold that startup to Palantir, which is where I got my startup start too in 2006.

as employee number 10. One of the things we used to talk about at Palantir all the time is everyone wants a button that you can press that would just make you money, that would just make you a decision. And the reality is that's not possible, nor should you want it, because it gets rid of the human intelligence part of the equation. In contrast, what Arena's doing is actually extending human capability, turning people into cyborgs. - It's so much about human augmentation versus sort of human replacement. If you're gonna be a cyborg, you need to control your cyborg suit or your Ironman suit and then you need to trust that it's going to do what it's done. And especially for something like this, it's so new, why would you trust it? And so a lot of the attention that we've put into building our products is to solve those two problems of control and trust. You could give as much or as little agency

to the machine as you want. So, I could just turn on discounts and say, All you can do is move discount between 50% and 70% off, or I could turn on discounts and bundles, let it time when they're going out, let it time how long the offer is, let it adjust in. I could give it as much or as little. So, that's the control piece. But then the second piece is you're handing over a set of decisions that maybe you weren't making before or were kind of getting made implicitly to a machine. So, how is it making decisions? And a huge part of this is understanding that. Well, what happens if, you know, we have a prediction the consumer's going to buy my product at that discount? What if they don't? So, what if they don't? Like, you want to see, what are you gonna do? You want some sort of trust that in an environment that's changing and dynamic,

how is your machine gonna respond? It's not enough in the business world to have something that will work only when things are on rails because, as we know, just living life, life goes off the rails all the time. (mellow music) - The F-22 is arguably the most advanced fighter jet in the world, but it can't be flown by a human without assistance due to its very design. It's so unstable that a flight computer must make 1,000 flight corrections per second to keep it in the air. There's still a pilot, but in many ways, the pilot tells the plane her intent for where to go or what to do versus exactly how to fly. - When you think about even the fighter jet example, you still have a pilot that's sending the intent of where they want to go. You have a flight computer that's making all the little decisions. They're extremely maneuverable, they're really agile,

and oftentimes the flight computer is actually making thousands of little flight corrections per second. So when you sort of joystick left, you're kind of sending the intent that you want to go left, but the little decisions required to implement it, they're autonomous. - The most powerful thing about running a company like an F-22 fighter jet is the ability to dive down to a specific customer and offer a hyper-personalized decision tree for every single customer. - That's the beauty of this, is at the top level, you get a summary. You see, oh, overall, like, what's my strategy? You can see I'm using, you know, quite a few just very shallow discounts, quite a few heavy discounts, not much in the middle, tend to dial down my heavy discounts over time. But then I think the beauty is really like, what is the right answer for each customer at each moment in time?

Recognizing that moments in time are different. It's not just, you know, tomorrow. It's tomorrow with heavy rain, or it's tomorrow and it's bright and sunny, or it's tomorrow and there's a really big game, and those are completely different situations. And you really want to get that feeling of a few different customers. And humans are so good at this, we're so good at that common sense, and it creates this really wonderful interplay with the human and the AI where, A, human starts to trust the machine, B, the machine gets better because it's being coached by a person. - The cool thing is this kind of thing has already been happening with AI across the board. GitHub Copilot has already been a literal co-programmer for legends like Karpathy. Here's a recent tweet where he says, I don't even really code. I prompt and I edit. Can you imagine this kind of technology fully realized

In every business and every business decision, a co-pilot that can be everywhere you can't? Arena's working on this. So, if you need this in your business, you can reach them at arena-ai.com. You can also follow Pratap on Twitter. Link in the description. He's an absolute expert and leader in this space. The future is so bright for software, especially one that extends human capabilities and turns us not into robots, but into cyborgs. Today was about highlighting one leader in the field, bringing groundbreaking techniques to businesses that could normally never get it. And literally, that's what an enterprise startup in the space is. This is all inevitable, but especially when building great enterprise software that works, it just happens a decade earlier. I can't wait to see what you build. I'll see you next week. (upbeat music) (upbeat music continues)

Applied AI for Pricing: You can’t be everywhere but your AI co-pilot can

Hi, I'm Garry Tan —I'm President & CEO of Y Combinator. I'm a designer, engineer, and investor in early stage startups. Previously Founder & Managing Partner of Initialized Capital, an early stage venture capital fund that was earliest in Coinbase and Instacart. 

Before that, I was a a partner at Y Combinator. Invested in and directly worked with over 700 companies the earliest possible stage, often just an idea. I cofounded Posterous and helped build it to a world-class website used by millions. (Acquired by Twitter) I also helped build the engineering team for Palantir Technology's quant finance analysis platform, and designed the current Palantir logo and wordmark.

I love building things. Forbes Midas List 2019 through 2022 🚀