Democratizing AI: Making Tools Accessible to All

Megan - 00:00:18: Today, my guest is Siqi Chen. Siqi is co-founder and CEO of Runway, the finance platform you don't hate. Runway empowers teams to build great companies by making business understandable and accessible to everyone. Companies like AngelList, SuperHuman, Stability AI, Lob, Grain, Sandbox VR, and many more, large and small, have chosen Runway to help them run their businesses. And more are joining them every day. Prior to Runway, Siqi was also the president, chief product officer, and CEO of Sandbox VR, the VP of growth at Postmates, and he sold his first company to Zynga in 2010. He is an angel investor who has invested in companies like Touch of Modern, Amplitude, Italic, and Clubhouse. He is a husband and the father of two daughters. Siqi, thank you very much for being my guest on today's episode of CFO Weekly.

Siqi - 00:01:50: Thanks for having me, Megan. Really excited to be here.

Megan - 00:01:53: Yeah, today our topic is artificial intelligence, which is obviously very hot right now, along with the changing expectations of the CFO. And I'm looking forward to learning about you and your experiences. So let's get started. First of all, I was just going to ask if you could give a brief overview of your career to date so that the audience has some idea of your interesting history.

Siqi - 00:02:18: Yeah, I have a somewhat eclectic background. I majored in math in college. I have a technical background, and most of my career has been as a founder of different companies. When I first got out of school, I was working for NASA JPL, first fell and then moved to the city. Worked for a search engine company called PowerSet, which was acquired by Microsoft and became Bing. I was basically an intern there. But right around then, I started a first company, and I was making games on social networks. We raised a few million dollars for Metric Capitalist, and that was acquired by Zynga. And I was head of product at Zynga through the IPO. And I left about two years later and started another company, which is its own story. The Washington Post covered it, but basically we were shut down by Congress for interesting and fun reasons that it's probably too deep to go in right now. But afterwards, that company was acquired by Postmates. And at Postmates, the VP of product at Gross, and I was there for about a year. And afterwards, I went to join a company that eventually became Sandbox VR. I joined as head of product, and then I was appointed CEO of this very interesting retail gaming hardware virtual reality company. We raised close to $100 million from Andreessen. I was appointed CEO of that company about six months ago. COVID hit, and then COVID hit. And we were scenario planning on spreadsheets with our CFO, really experienced, brilliant person. We hired from Netflix, and we had to hibernate the company throughout COVID, went from like 400 employees to 15 or so. I laid myself off as well and that's how Runway got started.

Megan - 00:03:54: Yeah, the founding of companies. What inspires a person to found a company? Just a great idea? Or have you always been an entrepreneur?

Siqi - 00:04:02: I think it's shifted as my career progressed. When I was younger, I felt like it was the only way for the world to give a person like me a fair shake. I didn't feel that what I thought I could do would be readily recognized if I were working in a big company. But the marketplace is unforgiving, right? If you can build something on your own, then the world rewards you for what you can build very, very directly. And so that was my initial motivation, is just like that felt like the fastest way to see what I can do. But now what motivates me is something really different. And what motivates me is just building something that you really, really want to see in the world. Something that when you look at the world, you're like, I can't believe this doesn't exist. And working on it for long enough and working on something where you feel like you and your team have a unique set of skills to finally make something possible. Something that wouldn't exist without you. I think that is like more motivating than any other incentive that I could have.

Megan - 00:05:05: Talk to us a bit about Runway. What is it that Runway does?

Siqi - 00:05:09: Runway is the closest thing that you could compare it to is FPMA platform. We do everything that you would expect in a traditional FP&A platform. You model in Runway. You can create scenarios. You can create reports. We integrate with all the different sources of data that actualizes your model. At a high level, it is FP&A platform. The thing that makes Runway different, though, is that we recognize all of these things are table stakes, right? You want integrations. You want modeling. You want dimensionality. You want ease of use. But the thing that makes runway is that when we think about how, and this connects to how the role of CFO has changed, right? When we talk to customers who are CFOs, the reality is that everyone knows that the role of CFO is increasingly more strategic. It's not just about the numbers. It's not just about control. It's not just about looking backwards, but it's about being strategic, being a real business partner, being a storyteller, and telling the story behind the numbers. And thus far, every vendor knows that this is the reality. And so what we've built as an industry, our tool to save the finance team time, thinking that if we can just build something that saves the finance team time, the CFO will have more time to be strategic and be collaborative. And the one thing that we believe very deeply is that saving the finance team time and building just for the finance leader is not enough. Because what real strategic requires is collaboration. And what collaboration requires is alignment. And you can't have alignment without having the entire team have the context necessary to get aligned. And even that isn't enough. You can like, Open up every spreadsheet which you know you can't, and you can share every source of data. But if the people in the company fundamentally do not understand what you're looking at, you can't get to understanding of the context. The context is useless. You can't get to alignment, which then makes collaboration and strategic impact really difficult. So at the core of Runway is we've designed an FP&A platform that makes the workings of the business and the plans around them understandable with clarity, through interactivity, through great design, to not just finance, but to the rest of the company as well.

Megan - 00:07:26: Yeah, wow, that sounds like an amazing product. So does it sit on everybody's laptop?

Siqi - 00:07:32: Yeah, so what's really interesting is, so for example, ConvertKit is one of our customers. And they did something really interesting recently that I haven't seen before. It did it all hands. And this all hands was centered around finance as a topic because most progressive companies, they want people to understand how the business works. And so the way they did that is they actually health scavenger hunt virtually inside runway. And have people look for different drivers and different metrics and discover how it works and find the issues around the formulas. And that's how they got their entire company really familiar with it. And they got a lot of great feedback from the team. So that is very much the idea. Finance is our first user and normally the primary user. But the idea is the great CFOs, as you know, know that finance is really not just about finance, right? It's about the whole widget. It's about everything. It's product, it's engineering, it's sales, it's marketing. And so in order for a team to really have a holistic single source of truth, you need to build something that is useful to the rest of the team too. So that they're not just having all these models and Google Sheets and Excel screwing around all over the place in Google Drive.

Megan - 00:08:49: And you mentioned having started your career with NASA. What lessons did you learn there that you've carried over and continue to resonate with you even in your current role today?

Siqi - 00:09:01: I was basically an intern at NASA.

Megan - 00:09:04: Amazing. I mean, anybody that can say that they worked for NASA. Is obviously brilliant.

Siqi - 00:09:10: It's amazing how, yeah, the value of a great team of people and people learn from them is definitely one thing I take away from that. That was like my first real job of any kind. I think the thing that it did for me is it gave me some confidence that I am capable of doing good work. Honestly, before this, I was not a great student. I didn't have a great GPA, but somehow someone at NASA took a chance on me and I ended up doing some really great work in machine vision for them. I'm actually the, I believe, the only person who's ever won a Congressional Space Act award while in school for the work I did on NER rovers. But before that, my parents growing up thought I was going to end up homeless because I was not the studious student at the time.

Megan - 00:09:55: And as the CEO of Runway, you've spearheaded the development of AI-driven tools to simplify financial analysis. So how do you see AI revolutionizing the role of the CFO, particularly in terms of decision-making and strategic planning?

Siqi - 00:10:12: We think about AI in a pretty specific way, right? Like it's very easy today to make a flashy demo around AI. What's harder is figuring out like what are the actual problems that a CFO has and how can we add value to the responsibility of CFO? And the mission of Runway, the mission statement is to make business accessible and understandable to everyone. And in a lot of the ways, you know, when you think about the role of a modern CFO, that has a lot of overlap with what a modern CFO does, right? Aligning the team, being a strategic partner to all the different functions of a company. And what makes it really inaccessible to people, as the joke is, like, the only person who really understands a model is one person made in the first place. The way we think about AI is in twofold. One, accessibility, and two, understandability, right? When you think about what LLM's large language models are good at today and what they're not good at. They're not good at doing that. They are very good at writing and having some sense of common sense and intuitive reasoning. With an FP&A platform, what you are doing is you're giving, effectively, this AI, this LLM, a calculator. And so the way you can make things understandable is you can look at any formula, you can look at any model or part of the model, you can look at any report, and you can have AI explain to you, in human language, what you're looking at. And you shouldn't have to chat with it. You shouldn't have to send a chat message for that to happen. It should just work in the background. When you look at something, there's just an AI isolation right there. And then that's very much how Runway works. So if you're looking at a driver or a formula, there's a plain English explanation of what this formula does and what is the source data of the formula that feeds into it. If you're looking at a variance, normally you have to drill in, you can see it, but we can just tell you, right? Like, oh, it looks like, your engineering expenses was 20% above forecast, and the greatest contributors to that is X, Y, and Z. That is something we get to know just in the background. That's how we make things more understandable. And in terms of accessibility, one of the toughest things is because if you don't understand the model, it's very hard to make changes to it. And so with AI, you can have it do the front work. You can ask Runway a question like, well, how about triple revenue next year and not run out of money? And it can look at the existing semantic structure of your model. It can look at, is there any plans? And come up with a reasonable guess and roughly fill those constraints. And of course, we give you the tools to easily fine tune it. But making things more accessible and understandable, that is fundamentally how we think about AI. And we want to express it in a way where you don't have to do a lot of work. You don't have to wait to chat with it or ask it a question. It's doing work for you in the background.

Megan - 00:13:01: And who is the ideal client as far as maybe revenue size or employee size? Who gets the most benefit out of a product like this?

Siqi - 00:13:11: Generally, we work with companies at the very small end. You're going to have around 50 full-time employees and at least like one full-time finance person, all the way to 1,000 to 2,000 full-time employees. That's what we define as a mid-market of the market. And so some of our customers today include, for example, AngelList runs on Runway today, Superhuman, the email company, Sable Diffusion, the AI company, Lambda AI company. We have a customer who's building literal space stations. So it's a very wide gamut of industries and sides of teams. But what connects them all is they are at a big enough size where the spreadsheets or a platform that doesn't allow for visibility across departments no longer serves their purposes and that's where we come in.

Megan - 00:13:57: And how is it that you envision AI assisting CFOs in identifying and capitalizing on either new revenue streams or maybe on the other side, cost savings opportunities within their organizations?

Siqi - 00:14:11: I think two of the biggest opportunities AI generally across like many domains, one is proactivity and the other is context. And these go hand in hand. To break that down a bit, if you think about ChatGPT, it's an incredibly useful tool. You know, we're all probably going to be familiar with it, but it is neither of those things. It is not proactive because you have to interact with it and it will respond to a message that you send, but it's not going to talk to you on a visible and deletion. So that's like one big opportunity. And the other is it doesn't really know much about you. It doesn't know what your role is unless you like enter some information about yourself in the system settings. It doesn't know like what your calendar looks like. What are your key priorities? What are the things that your team is doing? What are the sources of data that feed into your business and your model? And so the way in which AI can help a CFO do a better job and move faster is if you can give an LLM context around how the business works and what is important to you, if you want to strategically save money, what does the business even do? And AI can proactively look at the data and look at your existing plans and connect it with a context. It is effectively this can be this omniscient assistant that can see every detail of the business without you having to ask and proactively tell you where the opportunities are.

Megan - 00:15:39: Wow. That's really amazing. How do you teach it about the business? Do you just type that information in?

Siqi - 00:15:47: Yeah, I think that's generally the opportunity around finance in general, right? Like.

Megan - 00:15:52: Yeah.

Siqi - 00:15:53: This is not about just the numbers, not about a spreadsheet. It's about the story behind the numbers, right? What are people doing? The plans, the romance, the marketing campaigns, the sales forecasts. A lot of that context usually lives outside of the spreadsheet. And it is a big challenge to connect that important business intent and context with the members to make sense of it. And that is like one of the more important things that we try to figure out runway, which is how do we connect plans and intent with numbers? Because if we can show people the model and the plans in a way that maps to how an operator thinks of your business in a form of a roadmap, in a form of a plan, then it's likely to make it a lot more accessible. And if it's accessible, that means the whole team can plan more strategically. So yeah, we spend a lot of time trying to figure that one out.

Megan - 00:16:44: Yeah, that must be incredibly powerful. I know CFOs spend a lot of time trying to learn the business, especially when they're new to an industry. But how does that get into the platform? Do you just type it in or does it link to a company's website? How does the platform learn about the business?

Siqi - 00:17:01: Yeah, it is actually really hard to describe without just seeing the product.

Megan - 00:17:06: Yeah.

Siqi - 00:17:06: And the best way to describe it is like in the form of abstractions, like a spreadsheet of cells and rows and tables, right? And when you think about a business, it's very different from a sell, right? You think about these concepts like cash or pricing or growth or conversion rates. And so if we can build a product where the concepts, now slow searching how you think about a business, you can enable some more powerful workflows and make it more understandable. So for example, let's say you have a road called marketing spend, right? And if you imagine in your mind's eye what a spreadsheet looks like today, you can see like, okay, a million dollars in January, and you go right to the next cell, it's $1.1 million, and then the next cell is like $900,000, and so on. It's just this time series. And what we do at Runway is we build this thing where you can have plans impact the marketing spend. A marketing plan will look like this thing that you can move on a timeline. And it'll say, Q1 marketing plan, owned by Megan. And the impact of this plan is that we're going to increase marketing spend by 30K a month over the course of these three months. And also we're going to bump conversion rate by 5%. We're going to reduce pricing by 10%. And all of these are on the timeline. And it looks like a roadmap. And what you can do from that, is you can just drag that plan around. And it'll automatically impact your model in the right way. And if you think about one of the more annoying things to do with Excel is to change the timing of anything, right? You have to find the delta of what it was before and paste the delta, apply it to three months later, and then paste all formulas back and do it for every other row that you're trying to change the timing of. But with Runway, because plans is this, you literally just drag it in a second and everything will just work. Or you can just move the amount of marketing spend up and down visually. You can drag graphs up and down and it'll affect all the rest of your models. We really tried to build experience that was intensely visual, almost game-like, to help you give an intuitive visual sense of how the business works and how do your decisions impact it.

Megan - 00:19:16: And what do you see right now as the key challenges that CFOs face in adopting AI technologies? And how is it that they can overcome these challenges?

Siqi - 00:19:26: The biggest one is probably like, how do you make it secure? This is incredibly, incredibly sensitive data. And so I think a lot of the fear is around, well, how do we use AI in a responsible way where we're not leaking data to external services? I think that's probably a really big one. There's both on-prem solutions, but also, you know, with open AI enterprise, they're contractually obligated to not train on customer data, but that's certainly one source of friction we have, we've seen. And the other is just having an intuitive understanding of what ADI can even do. Like right now, most of the implementations of AI we've seen in the marketplace is really a chat interface where you can query it and answer questions. But you have to know what questions you want to ask and what questions it's capable of asking. It's very difficult to develop that habit. And so separating out and learning what are the things that AI is actually good at. And figuring out what are some implementations of AI that isn't just a flashy demo, but can actually add value. All of those are key challenges today because there is a lot of noise in the market.

Megan - 00:20:35: And CFOs are increasingly expected to provide strategic insights beyond just traditional financial reporting these days. How can AI empower CFOs to proactively identify market trends, forecast future performance, and contribute to long-term business growth strategies? I've even heard of some CFOs using AI to help them prepare for earnings calls.

Siqi - 00:20:59: Yeah, having a habit of using AI is generally a good thing. It's so much that you can build with AI, even without knowing how to code. And it was really surprised people. Looking at market trends, doing research, using perplexity to help you get competitive research are all great ways to use AI. But I would say this, actually, I think before AI even can add a ton of value to your business and into your work as a FFO, it's really important to have a good, I think the first step is figuring out how to give it context. Because right now, the thing that blocks CFOs from getting strategic impact, I wouldn't say it's necessarily enabled or blocked by the existence or non-existence of AI. I think what strategic impact requires is a deep understanding of every function of the business and have imbuing that understanding to the leaders of every function in the business. The reality is, like, we know every CFO wants to be strategic. And I think the blocker is actually getting the functional leaders of a company to care about finance. That is the reality, the lived experience a lot of CFOs is, you know, every quarter... They send out a spreadsheet to the functional leaders asking for a forecast and for a budget. And to the functional leaders, it just seems kind of like paperwork. That shouldn't be the case. And I think that apathy really comes from not being able to see or understand the business from the finance point of view. So I think the biggest blocker is communication and storytelling. Why does this matter? How does this business work? How does your decision impact the bottom line of the business and ultimately the value of your shares of the business? Those are the things that are really, really important. I think AI can do a lot to make all of that a lot more accessible to not just a CFO, but to everyone else in the business so that people value the role of finance and value the difference that a great finance leader can make. And one of our beliefs at Runway is that that is being held back by tools. If you're on a spreadsheet, for example, you don't share that spreadsheet. Why do you not share that spreadsheet? A couple of reasons. One is you can't share everything. There's probably a sheet full of employees and their salaries. And as you know, there is no function to just lock down one single column from everyone else and make that not visible and also allow everyone to see it and even if there were, you're not going to share it because someone might fat finger it. You don't want to give anyone else added access to this model because someone might change something. And those are problems that, frankly, existing products do solve, right? They solve problems around control, invisibility, and sharing. But once you have that, you still need to have a platform that people are willing to use, an interface where someone like a product person or salesperson come in and say, oh, here is our plan. Here is how my work and my roadmap as a product person or a salesperson, as a marketing person, connects to the impact that I have on the business. And those aren't things that exist today. I think the biggest blocker is communication and storytelling and helping the rest of the company understand the finance point of view of the business.

Megan - 00:24:16: And you mentioned that cybersecurity or AI being a cybersecurity risk, but is there a way you can envision CFOs leveraging AI technologies to enhance cybersecurity measures to protect sensitive data?

Siqi - 00:24:32: Yeah, I think this is an arms race. AI is a great source of leverage for everything that a company or a person could do. And that increases the leverage and power of cyber criminals, hackers. But it also means that it's an opportunity to increase the security of your data. But cybersecurity is far from my area of expertise. But that is what's going to happen. And you see this already happening. There's any number of cybersecurity companies are AI driven today, and they're going to be necessary for everyone, including the CFO.

Megan - 00:25:07: And you've touched on this a bit, but as the finance industry becomes increasingly interconnected, collaboration between CFOs and other C-suite executives becomes more and more crucial. So how do you envision AI fostering cross-functional collaboration and aligning financial goals with broader organizational objectives? I know tools like Runway help.

Siqi - 00:25:31: Yeah, I think it goes back to the combination of context and understandability. What large language models are good at today is English, translating really, really complicated concepts into forms that people can easily understand. For one example, I actually, if you go on ChatGPT today and go to ExploreGBTs, the number one educational GPT is created by Runway and myself. It's called Universal Primer. And you can ask it to teach you any enormously complex thing. And what this GPT will do is it will try to explain it almost like a five-year-old using really easy-to-understand analogies. You can ask it, how does a transformer architecture work in LLMs? Or how was Fermat's last theorem proven? And it'll try to give you a... Easy to understand explanation. But not only does it give you that, it then asks you, what is your level of understanding of all the prerequisite concepts necessary to understand the thing that you're asking about? If you're asking about how does like large language models work, right? AI. It'll ask you about your understanding of recurrent neural networks, your understanding of linear algebra, your understanding of attention networks. And depending on your level of understanding of each of those concepts, it will then recursively explain it and ask you how well you understand all the way down to arithmetic, if that's what's needed, so that you can build up all of the concepts necessary to understand something really complicated. And you can use the same exact idea generally across everything that happens in the business. You know, if a CFO wants to understand better how does a product work, how does a product work? And if the product wants to understand how their work impacts the CFO and the margin of the company and the growth rates of the company, AI is an amazing tool that allows them to learn on their own because they can just ask it a question or we show it to you and we tell you this is exactly how it works. I think the biggest thing that AI can do, and this cuts out a whole core of what Runway is about, is that we believe at the center of everything, of strategic impact, of collaboration. Between departments is understanding. And AI is very, very good at smoothing out the process of learning something, of understanding and getting everyone on the same page.

Megan - 00:27:49: And last question, but as you look ahead, what trends do you anticipate shaping the future of AI in finance? I know I've talked to a lot of CFOs who are just playing around with AI at the moment, but wondering what they can do to implement it or improve the business. Where is it heading?

Siqi - 00:28:08: Yeah, accessibility and understandability. I mean, I hate to just repeat our mission statement, but I think those are the two key things that AI will do more than anything to enable. This idea of not only just being a strategic partner to every department, but having the role of the CFO shift from like, here is the siloed monolithic function to this idea that the job of the CFO is increasingly to get everyone to have a finance mind, to have a business mind, to understand the trade-offs of decisions that they make, both in terms of expenses and the return on those expenses, on those investments. And AI is really, really good at that. It's really good at making things a lot more understandable. It's really good at making very complicated workflows. Like I want to figure out like a new sales strategy. I want to like model out a new product line. And you can just ask AI to do that. The democratization of finance and business thinking throughout the organization, I think is what AI will enable. And I think that is also dovetails nicely with how the role of the CFO is shifting from that one of command and control to context over control.

Megan - 00:29:21: Siqi, thank you so much for being my guest today.

Siqi - 00:29:23: Thank you for having me, Megan.

Megan - 00:29:25: Yeah, I'm quite sure some of our listeners are going to want to see the product. So how can they get a demo?

Siqi - 00:29:31: If you are a growing company and you have a full-time finance person or a hold up permit, and some of the pains that you're experiencing include you have way too many decision makers and it's not understandable to anyone else, or you have so much data as far to keep the model up to date, you can go to runway.com and you can see more about the product, look at our similar case studies around Superhuman and AngelList and Signover Demo.

Megan - 00:29:57: Thank you for sharing that. And thank you also for sharing your time to be here with us today.

Siqi - 00:30:03: Of course. Thanks for the opportunity, Megan.

Megan - 00:30:05: Yeah, I wish you and Runway all the best. And to our listeners, please tune in next week. And until then, take care.