Monday, 2:47 pm: Your CEO asks you to model 3 acquisition scenarios before Friday's board meeting.

In the old world, that’s 72 hours of spreadsheet wrangling and late nights. In Runway, you pull in actuals from your ledger, HRIS, and data warehouse - and build all three scenarios in 15 minutes. Complete with second-order effects on cash and hiring.

Runway gives you a live simulation of your business - so you can test strategies as you think of them.

See it in action.

Siqi Chen is the type of person we need more of in finance. A serial entrepreneur and AI maximalist. He thinks big and has had a fascinating career that started at NASA (as a literal rocket scientist).

He’s built and exited several tech companies and appears to be just getting started. He could have chosen any industry to disrupt next… but he chose FP&A software.

If you haven’t already, check out my first interview with Siqi from earlier this year. It’ll catch you up on how his winding path led him to Runway, which is turning the FP&A software game on its head.

When we spoke in January, we both agreed there was unfinished business. We needed to spend more time together looking ahead, diving into the future of finance, and understanding where AI was taking us (beyond the hyperbole).

So, we got together again and did just that. And now I’m excited to share our discussion with you:

SCFO: Hey Siqi. We last spoke about eight months ago. What’s new?

Siqi: There are two big changes: one is the internal changes here at Runway. We’ve doubled our headcount, quadrupled revenue, and signed customers like AG1, MyFitnessPal, and NBAT2.

But what's really changed is the pace of AI’s capabilities and the interest around it.

A year ago, we had AI features, but people didn't actually care that much. But now, when we sell, everyone wants the AI feature, everyone wants to hear about it. But it's almost to the point where they don't care what it actually does.

The reason they care about it is that the CEO has mandated that everything they buy must have AI in it. It’s a checkbox. It's really interesting how quickly that dynamic has changed over the past year or so.

SCFO: I was actually reading something today about a bunch of AI consulting engagements that have left F500 boards disappointed. Do you think there’s too much marketing bluster about just how “game-changing” AI will be in the near term for a finance teams?

Siqi: There’s a massive amount of marketing and hype there around all these tools.

What I found is that AI adoption is actually quite high, generally, in finance. It's just that the particular type of AI they're all using is ChatGPT. And my takeaway from that is when there is something truly valuable and useful, adoption happens very rapidly. People opt into it.

It will unquestionably get better from both a model capabilities perspective and an applications perspective. And when they cross that threshold, people are going to be lining up at the doors very quickly.

SCFO: As a CFO, part of your comfort in the numbers your team is producing, whether actual reported performance or forecasts/scenarios, comes from the process itself. The relationships between the numbers.

Internalizing those relationships helps you build a kind of warm, fuzzy feeling about the numbers you are putting your name to. How do we maintain that intimacy with the numbers and the business in a world where we could, in theory, have an all-knowing reporting AI bot at our disposal that could just spit out the right answer?

Siqi: The answer is useful, but it's even more useful to understand how you got to the answer. Understand the relationships, and develop this intuition, this feeling of “I understand how the business works.”

What this speaks to is what finance is actually about.

There are two parts of finance: forward-looking and backward-looking.

Looking back is reporting. But the more strategic bit of finance is “finance as thinking.”

It's not about getting an answer. It's about developing intuition. That warm, fuzzy feeling you're talking about, where you actually understand the drivers of a business.

And so the purpose of building the model is not just getting the answer. The value is actually in building it.

In combat terms, the value of a plan isn't the plan itself, right? It’s going to go out the window when you get punched in the face. But it's the act of planning that’s valuable. And it’s the same for modeling in finance.

SCFO: So, do you see AI as an accelerant or an oracle for “finance as thinking”?

Siqi: We deeply believe that finance at its most strategic will always have a human element.

An analogy is, instead of AI finance tools like Runway being a godlike Terminator thing, what if it felt more like an Iron Man suit?

It helps you make the complexity of the business more understandable so that you can develop a better intuition of that complexity yourself [Iron Man suit] vs. an all-knowing overlord that simply gives answers and makes your team redundant [Terminator].

And when you have ideas, or thoughts about what you want to do, it can accelerate the time to offload that intent and understand the consequences of it so that you can play with intuition and develop an answer. Not simply get the answer.

That's the difference between a tool for thinking that is more like Excel and something more rigid that is your typical FP&A platform.

AI is really an accelerant to that. It's not telling you the answer. It makes things more clear, it helps you think faster. That's very much our philosophy around how AI should be used.

SCFO: Last time we spoke, you said “everything is finance,” which I took as a call to arms for CFOs to figure out how to externalize finance thinking.

And I know that idea is core to Runway’s mission.

So if Runway succeeds, does that effectively make the business self-service on ‘finance’? To me, it sounds more like that Iron Man suit is ultimately for the business, not the finance function.

Siqi: Possibly. Good question. I think it certainly does some of that.

It democratizes access to understanding the businesses finances to everyone in the company. And that is something that a really good strategic finance person, finance org, actively works on, right?

You have embedded finance partners to make sure that there is a bridge between the two. And so I think in the short term to midterm, building better tools and technology to make that better is something that everyone wants. But at a certain point, if the technology gets so good, then what does a finance person even do anymore?

What is their role? Is it just data management? Are they the CEO now? And if you can be the CEO, then why do you have a finance team?

Because maybe the CEO can just have this Iron Man suit, and you don't need a finance team anymore. 

And I'm not going to say that's not possible.

But I will also say that in that world, that would be a true statement about every other function. If AI helps you design a product, why do you need product people, and why do you need designers?

And I can speculate in some depth on what that world looks like, but I think that is not at all specific to finance.

That's just a societal, “what do we do with AGI?” type of question.

SCFO: Excel has ruled in finance because of its flexibility. But there’s a point where its weaknesses like workflow management, data connection, version control, etc. become a problem. That’s normally when teams invest in FP&A software. Traditional FP&A software does a good job of solving those problems, but at great cost (not just the $)… you lose flexibility.

I would assume you are making product and feature decisions every day that are trading off between a traditional FP&A platform with the flexibility of a Google Sheets or an Excel. How do you make those decisions?

Siqi: We think about it like Legos.

Legos are a form of emergence in that you have a Lego set, and you can build all kinds of different things. And certain kinds of software are like that. Airtable, Notion, Figma, Coda, they have that feeling to it.

You can build software in a way where it solves those problems, and it'll solve problems that you haven't even thought about yet.

If we build Lego blocks that are so well-designed and so well thought through, customers can get the power of a traditional FP&A tool and all the flexibility.

SCFO: One of the barriers to AI adoption I hear consistently is the privacy and compliance concern piece.

Siqi: 20 years ago, cloud and web software became a thing, right? Everything was on-prem, and Salesforce had to basically invent and evangelize the idea that actually, it's okay for you to have your very private customer data in the cloud, in the multi-tenant environment.

That was scary. Until it wasn’t. It’ll be the same with AI.

SCFO: Talk to me about the ‘hallucination rate’ with AI. I think to me that is probably the second-biggest barrier to direct AI adoption for high precision work (like finance, legal, etc.) in enterprise teams, at least.

Serious boards carrying real personal and professional risk, have a very low tolerance level for words like ‘hallucination’

Siqi: There are many, many reasons to criticize LLMs and AI capabilities. But I'm struck by how common and frequently there is no accounting for the progress of reasoning capabilities and hallucination rates.

I don't think it'll go down to zero because AI is a lot like people. We will guess, and we will make things up, and we'll be wrong about things that we're not completely sure of.

SCFO: So it's effectively a transitory issue? Is that the way you think about it?

Siqi: The thing about AI is that when people hear it, they think, “oh, this is very complicated, science fiction technology.” It sounds really scary.

But the reason why it is so powerful, so capable, so intelligent, and so useful, is that it is actually more like us and our brains and the way we think than software used to be.

That means it's going to be good at the same things that we are. But, it's also bad in the same ways that our human brains are bad. We are not good at doing mental math. And so if you were to ask a finance person to solve a problem, they load it into a software tool, and the tool calculates it in the form of a spreadsheet.

Now you can ask an LLM that question, and the right way for it to answer that question is to know how to use the tool.

SCFO: I’m not so sure human errors and AI hallucinations are the same. For two reasons.

Firstly, my personal experience with AI hallucination with numbers is that LLMs are much better liars than humans. The LLMs are very convincing when they are making stuff up.

And secondly, the possibility of a ‘hallucination’ in, for example, a reporting task, breaks the CFO’s mental ‘audit trail’ behind how something is put together.

Hallucination introduces a variable that the existing processes and teams aren’t built to control for. Error tolerance is going to be incredibly low in accounting and reporting functions. That could, and will, change over time as teams adapt, but that is going to take a while.

Siqi: I think AI has a branding problem in that it's not really hallucinating, it's kind of bullshitting.

It’s like a person that might just make up an answer if it doesn’t have the right tools and the right prompt. At least you can prompt it to say “I’m not sure” if it’s truly not sure. And I think that points to a prompting issue.

As for the question, “how do you know if it's right?” I’d turn around and ask, “how do you know the people who are working for you building models are right?”

You vet their work, right? You would check [AI] the same way. And over time, the need to check AI’s work goes down.

SCFO: Let me pitch two possible future states for finance and the CFO domain.

First, the bull case. It’s the fashionable view. As the CFO gets to automate manual tasks, the ‘people’ in the team can focus on being more strategic and building the bridge from finance into the rest of the business.

And as that happens, the CFO domain could become the natural home for adjacent functions, like strategic insights, operations, etc. We’ve seen some of that already.

The newly created COFO role at Salesforce, for example. While CFO and COO hybrid roles are not new (I’ve done one myself), it normally looks more like an exec double-hatting, or it being interim in some way. Memorializing it under the COFO title - as happened in the Salesforce example - feels significant to me.

So that’s the bull case - a super CFO who becomes so effective they gobble up adjacent functions.

But then there’s the bear case for finance, which I don’t really hear anyone talking about…

Here’s how it could go.

The controlling function (accounting and reporting) is going to get ripped to shreds by AI and automation over time. Why wouldn’t it? And at the end there will likely be this ‘super controller’ who runs what used to be a team of 50 with now just a bunch of bots and a handful of humans. That super controller will run a highly efficient, tech-supercharged compliance and reporting function.

So could there be a world where the strategic parts of finance get swallowed up by an operations function? Or it becomes embedded into the individual functions via intelligent assistants, powered by the information produced by the ‘super controller’? And the definition of the CFO role kind of gets ‘relegated’ to where it was 20+ years ago… compliance. Just a tech super powered version.

I see the traditional ‘finance’ function heading in one of those two directions, and probably not much in between. What do you think?

Siqi: One bookend is certainly that finance becomes more strategic, more operational. The COO and CFO start merging. Not a new idea.

The other one is finance is just completely focused on the finances. The traditional finance things, except just massively automated, whether it's reporting, accounting, modeling, or forecasting.

But, if I were to bet money on it, I would bet that those two outcomes plus a third one will all happen, depending on the company.

They don't seem exclusive to me. Actually, they all seem like almost the same thing, and I'll explain what I mean by that.

If you think about the role of CEO, you have to mentally merge all the different functions. Your role is literally operations and finance and product and engineering and all these different things. Technology and tools make everyone more cross-functional.

As a CEO, it is useful to look at things through multiple lenses. So there's an engineering lens, science lens, product lens, marketing lens, the operations lens, and a finance lens. And not everyone is going to be good at switching between all the lenses.

And so I can also see a world where, yes, it is far more democratized, but there's still enormous value in having explicitly a finance lens.

So that's the third outcome that I'm alluding to, which is that finance is more strategic, more embedded than ever.

And I think that, depending on your culture, you could have any one of those three. I think we'll see all of them.

SCFO: So given that … where the hell does the next generation of finance talent come from? And how do they add value?

Siqi: My view is that craft and proof of human and what is provably scarce all become more valuable over time.

I think that's going to be true across every discipline, every function. And so what are the creative acts of finance that are sort of subjective and opinion-based? I think a lot of that has to do with strategy. And the things that are objective will all be automated.

SCFO: I am generally not an advocate of finance pros vibe coding new software solutions to solve finance problems. We are fortunate to have some of the best software engineers in the world working on the next-gen finance stack, powered by AI. Isn’t the best way to deploy AI into your business to just pick the best tools, embed them in your business, and grow with them?

I see finance pros vibe coding as a good way to experiment and understand the capabilities, but why should an FP&A analyst think they can create anything in-house that is going to out-engineer what you are building at Runway, for example?

Siqi: I think in this environment when there is so much change and so much hype, it is a very good strategy to ‘build to learn what to buy.’

Having an understanding of how products work, what it's capable of gives you a much better intuition when you figure out what the right vendor is to buy or to not buy from because you understand maybe it's not quite there yet.

So I think there's utility there. There's also utility in just future-proofing. If you're the person who actually understands AI at a deeper level, then you're going to be a lot more valuable than someone who's just doing the accounting all day.

Another reason why it [vibe coding] still might make sense in the short term is that you get to build exactly the thing that you want. Being able to do things exactly the way you want is valuable, at least in the short term.

In the midterm, the tools that you're going to have are going to be so good that you can just buy them, and they will support all these different edge cases.

But then, over the long run, you can imagine a world where actually vibe-coding tools get so powerful that they could build something as good as any SaaS company, if not better.

And in that world, things are very weird.

SCFO: That’s a great way to think about it: “Build to learn, what to buy”

When we stop to look back on the last 30 years, I think it will be defined as the “ERP Era.” A period remembered as a time of overpromising and underdelivering on the part of the ERP industrial complex (software providers, consultancies, and implementors). During those 3 decades or so, far more value was created for ERP software providers and consultants than for the businesses they claim to serve.

What excites me most about the AI era is that it people like Siqi will tear up the traditional finance tech stack, and maybe, just maybe, we can finally get to the technology that works like “magic.”

Still modeling like it's 1999?

Back then, we had Excel. Today, we still have Excel.

But now your CEO wants real-time scenarios. Your board wants instant implications. You spend weekends rebuilding models because your team changed the roadmap ahead of the board meeting.

Meanwhile, teams using Runway are modeling live. They change one input and watch it ripple through cash, runway, and hiring - instantly. Start thinking out loud about the future of your business.

See what financial modeling actually looks like in 2025.

If you enjoyed today’s content, don’t forget to subscribe.

Disclaimer: I am not your accountant, tax advisor, lawyer, CFO, director, or friend. Well, maybe I’m your friend, but I am not any of those other things. Everything I publish represents my opinions only, not advice. Running the finances for a company is serious business, and you should take the proper advice you need.