Expensify Inc (EXFY) 2025 Q3 法說會逐字稿

Welcome to the Expensify Q3 2025 earnings. I'm CFO, Ryan Schaffer. And with me I have our Founder and CEO, David Barrett. And now I'm going to hand it over to Niki for the legalese.

Please note that all the information presented on today's call is unaudited. And during the course of this call, management may make forward-looking statements within the meaning of the federal securities laws. These statements are based on management's current expectations and beliefs and involve risks and uncertainties that could cause actual results to differ materially from those described in forward-looking statements.

Forward-looking statements in the earnings release that we issued today, along with the comments on this call, are made only as of today and will not be updated as actual events unfold. Please refer to today's press release and our filings with the SEC for a detailed discussion of the risks that could cause actual results to differ materially from those expressed or implied in any forward-looking statements made today.

Please also note that on today's call, management will refer to certain non-GAAP financial measures. While we believe these non-GAAP financial measures provide useful information for investors, the presentation of this information is not intended to be considered in isolation or as a substitute for the financial information presented in accordance with GAAP. Please refer to today's press release or the investor presentation for a reconciliation of these non-GAAP financial measures to their most comparable GAAP measures.

Thanks, Niki. Now let's dive into the Q3 financials. Revenue was $35.1 million. Average paid members was 642,000, and total interchange was $5.4 million.

Our operating cash flow was $4.2 million. Our free cash flow was $1.2 million. Net loss was $2.3 million. Our non-GAAP net income was $4.3 million. And our adjusted EBITDA was $6.5 million. Q3 free cash flow was a little less than in prior quarters. That's mostly due to seasonal timing of some annual payments. We also reiterate our fiscal year 2025 free cash flow guidance of $19 million to $23 million. As always, here's our Q4 flash numbers for our paid members in October. Up from the Q3 average, which we always like to see, 653,000.

And now to jump to some business highlights for Q3. We had some great marquee customer wins. We are now the official travel and expense partner of the Brooklyn Nets, who's a long-time customer of our expense product, and they have adopted expense by travel that shows just the power of the platform and the fact that customers are really excited about this. So we're very happy to have the Brooklyn Nets as a new expense by travel customer.

On the topic of travel, bookings continue to climb, growing 36% from Q2 and 95% since Q1, so expense-side travel continues to be a bright spot in the business and something both us internally and our customers are very excited about. We also repurchased 1.5 interchange shares of our Class A common stock, and that totaled approximately $3 million.

And now I will hand it over to David for a product update.

Great. It has been an extremely exciting quarter when it comes to the product side. First off, talking about migration, as you know, everything hinges upon our ability to move existing customers over to new Expensify. That's what triggers, we think everything in the business is recovery and growth and so forth. And so we've made incredible progress in that.

At this point, we would say we're targeting what we call 90% feature parity, meaning we want to support essentially 90% of the functionality of classic on new Expensify. We're very close to that right now. Of course, we'll always maintain classic for existing customers as long as they need it, but the main thing right now is that new Expensify is largely complete when it comes to the functionality of classic. We've also migrated basically the data of nearly all customers to new Expensify, meaning that customers can switch back and forth between new and classic as they like, which is a huge accomplishment.

So we're to the point where essentially new Expensify is essentially done from a feature perspective, and now we're just carefully what we call nudging customers over, meaning that we will make them sign into new Expensify the next time they sign in, but then they can optionally switch back to classic. We've nudged all of our collect customers over.

Now, to be clear, we have basically two plans, collect and control. And so our collect customers are smaller, simpler customers. We've migrated nearly all of them over to new Expensify, and the vast majority choose to stay on new Expensify rather than going back to classic. And so this is a huge testament to the power of new Expensify. Additionally, and I'd say this is one of the most exciting things.

Now, we're closing all new customers on new Expensify, meaning that we will start every sales conversation on new Expensify, and we'll still switch back to classic if there's some long-tail features, some esoteric integration, or something like this that they might need. But we start every new conversation on new Expensify, and so that's been really, really powerful, especially at the conferences, especially as we roll out to new leads. So it's been great, great progress when it talks about migrating existing customers from classic to new.

Additionally, it's been very exciting on the Concierge side. We've been talking about this for a while. If you've been paying attention, AI is kind of a big deal. We've been talking about AI for a long time because new Expensify's entire design anticipates basically modern AI.

And the way that we view it is AI is incredible, but it's also not foolproof. And so whereas some people really focus on AI in absolute, we view AI as a great feature for certain levels of functionality, and we would take the AI as far as it can and then have humans take it the rest of the way.

And so our design, which is very unique, it's a hybrid system. When you talk to Concierge, if it's a simple, common question or even just something very detailed about the product and sort of like from a help page or whatever it might be, the AI is really great at handling that question. It can do it better than a human, honestly. But if you get to a super complicated topic for diagnosis or if you have more of kind of an emotional issue, that's where we bring in our human agents.

Now, we can seamlessly switch back and forth between AI and humans sort of imperceptibly to the customer. And so to the customer, all they get is just an incredible chat support experience. But on our side, it's handled using AI or humans seamlessly depending on who's best for the job. Likewise, this is a contextual AI, meaning that it's built into the product rather than sort of on top of the product.

I think you've seen a lot of AI solutions which are kind of like Windows 95 Clippy, where basically it's just something kind of stuck on top. It's very clearly not designed around the product. Ours is different. With Concierge, it's built into the product in every place. And so wherever it's natural for you to talk about -- talk to the AI, whether either you're talking directly to Concierge, or maybe you're inside of an expense report, or even commenting on a particular expense, our AI appears everywhere. So you can basically talk to it naturally in the context of that.

Additionally, we're building what I would call a general intelligence. I think there's a lot of different approaches towards this, and the most straightforward approach that people start with is they'll have kind of a collection of very purpose-built agents, and so maybe a specific agent will reach out to you in a particular narrow context and talk about one topic. It makes sense. That's a very easy place to start, and I think that's kind of where everyone starts.

Our design is going for more of a general intelligence, meaning that we've built a singular AI that can operate in a multimodal fashion. So you can talk to the same AI and you can ask it to scan receipt, categorize an expense. You can ask it very complex questions about how to configure Expensify. And so the same AI can do all of these different functions.

What's nice about that is it really supports our contextual design. So it's not like you have to have 10 different AIs hanging out in every single context, and then you have to choose the right one based upon the question that you have. Rather, you can send any question to Concierge, and they'll always be able to answer it. This works especially well across platforms. So you can talk to our Concierge sort of like single general intelligence over chat, obviously, but you can just email it at concierge@expensify.com or just text it 47777. And because it's a single general intelligence, you can ask it any questions in any of those. And so you can ask it to create expenses, ask it about your expenses, about your workspace, whatever it might be.

This is a really powerful platform that we think is unique and novel in the market. We don't think anyone else has this level of sort of general-purpose financial AI out there, and this is just a start. To give some examples of how this works in practice. So there's some basic stuff. Of course, obviously, detecting not just whether the expense from the merchant and amount is out of policy, but looking into the receipt itself, making assessments about what type of merchant it is, and so forth. And so we can do more detailed prohibitive expense detection.

Likewise, it's all the rage these days. AI is a big deal for not just the admins, but also for the employees. And so we detect AI-generated receipts and flag them. We have a feature they call conversational corrections, meaning, of course, whenever you swipe the card or scan a receipt, we will categorize it to the best of our ability based upon the information just on the receipt and merchant itself. But every company is different, and sometimes there's ambiguity as to the correct way to categorize it.

So we'll narrow it down to a short list of the most likely options and just ask you, which one is it? If you're in the app, you can just do it in one tap. If you're responding via text or email, you can just respond with a number or whatever. And you don't have to pick from these options. You can also just say something else entirely. This is the advantage of a general AI, where if it asks you a question, you're not trapped into whatever conversation it wants you to do. You could actually just switch the script and ask maybe like, well, what are all the categories available? Or what's the last time that I did this? Whatever it might be. And so this general intelligence allows for a much more natural ability to correct and sort of categorize information.

And as mentioned, this is a truly universal agent. You can have the same conversation in a wide variety of contexts, whether it's chat, email, SMS, and so forth. So this is a major release for Concierge AI, but it's really just the start. We think this is an incredibly powerful foundation that we've ironed out the kinks for, and you're going to see more and more incredibly powerful functionality being built across it over the quarters to come.

So just to kind of summarize everything at a high level, we've increasingly and continued selling in a very successful fashion travel and card to existing customers, which has been great. We've been putting our free cash flow to work, which is great. And despite all this, despite all the chaos of everything, we've really stayed focused on investing in an AI-first design. And I think this is a big deal because obviously everyone thinks a lot about AI, but I think that everyone's kind of gotten through the first wave, a lot of the easy stuff.

Here's where it starts to get much harder going on now. And so we think that chat is the UI of AI. If you can't talk to it, how smart can it really be? And so our design is to bring a chat-first design everywhere into the product such that it makes our entire product into an AI-first design. It's a very, very different design. I'd encourage you to check it out, and I think you'll see a glimpse of the future, because we think everyone's going to be designing something like this over time.

Likewise, our new Expensify migration is on track, and we've got really great customer reception. This puts everyone into a position to talk with their AI in a much better way than they could with their previous product. And at the end of the day, it's really about anything that you can do via the UI, you should be able to do via AI. And so building a truly AI-first product where you can talk to the AI in a primary mechanism as opposed to just as a sort of secondary flow.

Anyway, we're going to have lots more to talk about in the quarters to come, but for now, we'll just take any questions we can.

Perfect. Let's get started with Citi. I believe George, you're on the line.

Great, thanks for taking the questions here. I'm on for Steve Enders. Maybe just on this point about chat as the UI for AI, this is something that you guys have been early to -- it's interesting from our perspective to watch other people kind of catch up to where you guys are in terms of building in natural language-driven UI into other software apps.

I'm just curious, from that head start that you've had, what have been some of the big learnings or capabilities you've incorporated into the platform that when you watch others, you can see maybe them making missteps or where you feel like you have an advantage there?

That's a great question, and I think it really comes back to this idea of being built in versus built onto the product. In that expensive ad design is that you can go into any context, and inside that context, you can talk with your AI about that particular thing.

Now, this is kind of a nuanced point. But imagine, for example, you're texting with an AI in a general context, and you want to change yesterday's expense. You want to basically categorize it, or you want to highlight that actually that was an accident. I didn't mean to submit that, whatever it might be. Referencing that outside of the context is actually quite hard. You have to remember the merchant, the date, the amount, or some key indication of how to do it. And it's just a really impractical thing. That's going to drive you back to the UI.

Now, if you're talking to your assistant, you would just say, hey, that thing that I did yesterday, or whatever it might be, and you'd give kind of a relative reference, and it would be able to figure it out based upon the contextual clues of the conversation. And so I think that our UI is about trying to infuse the AI throughout the entire product, such that you can use it in whatever context you're already in. You don't have to leave your context to use the AI. It's already there.

This makes a very different UI design. You can see it's a very chat-centric design. In many ways, it looks like a kind of ChatGPT interface. I mean, I think that it's hard to argue your business is an AI-first product if it looks like Concur. I think that if it had to look a lot more like ChatGPT to really, you know, credibly say that this is an AI-based thing.

It's kind of like, what makes an AI intelligent isn't that it just has a bunch of kind of AI branding on a bunch of algorithms. I think you need to be able to talk to it. You have to be able to ask it questions, whatever you want. You have to explain why it did what it did, and you have to be able to learn from its mistakes. I think that the idea that you can have automation in place and that you can't talk to it and figure it out doesn't seem very smart. Let's say you hired some sort of an accountant, and they said that they approved a report, and you asked them, why did you approve the report? And it's like, I don't know. You wouldn't be like, this is a genius. You'd be like, this is pretty stupid.

I think a lot of algorithmic automation is very powerful, but it's not intelligent in an AI sense. I think intelligence is about getting into a place where you can ask questions, get answers, and make changes all through natural language, and I think our design is really optimized for that.

I also think it's important that you're unlocking a new use case. Making charts with AI is not interesting. But that's a very common use case. People have been making charts for a long time and it doesn't require AI. That's not a good use of AI.

So I think the fact that we're able to do new things, new functionality, offer new value to the user because we're using AI is what sets us apart versus replacing code with AI. The customer doesn't care about that.

Super interesting, I appreciate the detailed answer there. Maybe something more tactical, the government shutdowns in the news, it seems like maybe there might be some impact on travel. I can appreciate that probably if there is any impact to you guys, it would basically be a timing risk, but just any thoughts there from shutdowns in the past or just general scenario analysis you guys maybe have thought through there?

So I think it's -- I guess it depends on to the extent it impacts travelers. If you're stuck somewhere, you're probably going to actually end up spending more because you have extra hotel nights because you're stuck in New York or something.

But in terms of is it going to keep people from using expense if I travel less or something because they're worried of being stuck, I think that's probably a realistic risk. It depends on whether people are going to change their travel plans or just risk it, basically, I think.

Yeah. I don't think uncertainty is good for anyone's business.

Yeah. I appreciate that. I appreciate you taking the questions.

Great. Let's see. JMP, I believe Aaron, you're on the line.

I'm here. Thanks, guys.

Hey, Aaron.

Hey, Ryan. I want to dig in on migrations from Expensify classic to new Expensify, including what percentage of revenue today is on new Expensify after migrating your collect customers, and the timeframe over which you expect to get your control customers that I think are a substantial majority of your revenue on new Expensify.

That's a good question. I don't think we know the -- it's less than 50% of revenue, so we're not over the 50% hump in terms of revenue yet, but that's the huge priority right now is moving people over.

Yeah. I mean, as I mentioned earlier, we're aiming to have new Expensify match classic from a functionality perspective by end of the year, and I think we're very good on that target.

Now, the real question is, how fast can we migrate everyone over? We control the timeline here. There's no sense migrating them over faster than they're comfortable with, and so we're going at the fastest rate that they're comfortable with. I think we're really hoping to have a significant progress on that, if not completion or near completion by the end of the year, but I don't think we can control. We don't know exactly yet because we don't know if we don't know.

Yeah. We're also listening to the feedback of customers nudged and iterating very quickly because people who are new to Expensify and they come in, they love it because it's all they know, it works great, it's very cool. Someone switching from who's used classic for maybe 5 years, 10 years to new, that's a different audience and they have a different reaction. It's not negative, but they have a different set of feedback than what we've gotten just from new customers coming in.

So we've been a little slow moving people over and really focusing on those user sessions and getting feedback and making small changes quickly and iterating. And I think it's flywheel where it goes faster and faster, but the existing customers are an interesting source of feedback compared to net new because they say different things. So we're just working through that.

Actually, that's a great point. The bottom of the slide, it talks about a major goal we had was to make sure every new customer conversation started on new Expensify. And so that has been the priority. That's done. And so now the priority is getting existing customers over.

That makes sense. And then the follow-up here, are you seeing any incremental monetization from the customers that have migrated to new Expensify or is that more a TBD? And I assume the more relevant piece of this question at this time is what type of internal cost savings do you anticipate from the Concierge agent once you get everyone migrated over to new Expensify?

That's a great question. So the support cost should be definitely less when we get everyone over because new Expensify handles everything better than classic. A lot of the problems -- not problems, but there are some complaints with classic that we have solved with new Expensify, so in general, it should be less of a support burden. Also, just the fact of maintaining two platforms at once is expensive and can be a split-brain problem, so it'll be -- we're really looking forward to solving that.

In terms of increased monetization, I think it's much easier to issue new cards, manage everything, get into travel. There's a lot of travel functionality that only exists on new Expensify. Using Expensify travel with new Expensify is a better experience than classic. I do think that it's a net positive. Everyone that we move over is a net positive on the business, so that's why it's a huge focus for us right now.

Absolutely. All right. Thank you, guys.

All right. We were double booked with some of our other analysts, so we will talk to them offline. That's everybody for now.

Great. All right, thank you all, and we'll see you next quarter.

Thanks, everyone.