Latest / The Dutch Investors / #110 | AI Winners or Victims? The Case for Klaviyo, HubSpot, and Adyen

Transcript

#110 | AI Winners or Victims? The Case for Klaviyo, HubSpot, and Adyen · The Dutch Investors

• 0:00
  SPEAKER_00: Hey everybody, welcome back to the Dutch Investors Podcast.
• 0:03
  SPEAKER_00: Today, AI victims or AI winners.
• 0:06
  SPEAKER_00: I think this will be a fun episode because we've been getting tons of questions about what's going on with all the AI disrupting software companies, so we will discuss this more in depth.
• 0:19
  SPEAKER_00: Before we continue, I don't usually ask this, but it would really mean a lot to us if you could give us a five-star review on your favorite podcast player.
• 0:28
  SPEAKER_00: It's the best way to help us grow and get this podcast visible for as many listeners as possible.
• 0:34
  SPEAKER_00: It's totally free, takes a few seconds, so I appreciate it.
• 0:39
  SPEAKER_00: Let's get into today's episode.
• 1:01
  SPEAKER_00: At first, you check the news, the socials, you start checking the results.
• 1:07
  SPEAKER_00: You see that revenue is still growing.
• 1:09
  SPEAKER_00: Customers are still signing up.
• 1:11
  SPEAKER_00: So you tell yourself that the market is being irrational.
• 1:15
  SPEAKER_00: Right?
• 1:16
  SPEAKER_00: Then the price falls another 20%.
• 1:20
  SPEAKER_00: It scares you a bit, and you start checking everything again.
• 1:24
  SPEAKER_00: We think quite a few software investors know that feeling by now.
• 1:28
  SPEAKER_00: The explanation they keep hearing is AI.
• 1:31
  SPEAKER_00: Companies will build their own software.
• 1:33
  SPEAKER_00: AI can now reverse engineer almost any software program, from Adobe to Microsoft.
• 1:41
  SPEAKER_00: Employees will be replaced by agents.
• 1:44
  SPEAKER_00: Businesses will need fewer subscriptions and seats.
• 1:48
  SPEAKER_00: Some of that concern makes sense to us and it's reasonable.
• 1:53
  SPEAKER_00: We think many of those assumptions might actually turn out to be true as well.
• 1:58
  SPEAKER_00: If software becomes much cheaper to build, competition will change, and if a company charges for every employee who uses its product, fewer employees could mean less revenue.
• 2:10
  SPEAKER_00: Now today let's look at Clavio, HubSpot, and Argen.
• 2:14
  SPEAKER_00: We find it difficult to put them all under the same heading.
• 2:17
  SPEAKER_00: What customers pay for them is quite different, and the business model for these three companies are quite different.
• 2:25
  SPEAKER_00: The shares of these three companies have fallen heavily.
• 2:28
  SPEAKER_00: But we think there is a reasonable argument that AI could actually help these businesses.
• 2:34
  SPEAKER_00: The question is how much of that argument survives a closer look.
• 2:38
  SPEAKER_00: Clavio, for example, was down roughly 65% from its 2025 closing peak.
• 2:45
  SPEAKER_00: HubSpot was down around 72%, and Argen around 52% from highs.
• 2:51
  SPEAKER_00: When you see a business growing, while the share price has more than halved, and naturally you want to know what you're missing.
• 3:00
  SPEAKER_00: Perhaps other investors have spotted a problem that has not appeared in the results yet.
• 3:05
  SPEAKER_00: Perhaps they are expecting a problem that will turn out to be much smaller than they fear.
• 3:10
  SPEAKER_00: Let's start with Clavio, because you have probably encountered its product without ever knowing it.
• 3:17
  SPEAKER_00: You leave something in an online shopping basket, and a little later an email arrives reminding you about it.
• 3:23
  SPEAKER_00: Or you buy a jacket, and a few weeks later the shop suggests something to wear with the jacket.
• 3:30
  SPEAKER_00: From your side it's just another email, but from the shop side, there's quite a lot going on.
• 3:36
  SPEAKER_00: It needs to know what you bought, what you looked at, and whether you have already returned something.
• 3:42
  SPEAKER_00: It has to decide who should receive the message, then work out whether sending it was worthwhile.
• 3:49
  SPEAKER_00: Clavio brings together the customer information and helps businesses use it in their marketing.
• 3:55
  SPEAKER_00: You might look at it and think, surely AI can do all of this as well.
• 4:00
  SPEAKER_00: Parts of it should become much easier.
• 4:03
  SPEAKER_00: Writing the email is an obvious example.
• 4:06
  SPEAKER_00: We would be surprised if producing decent promotional copy remains expensive for very long.
• 4:12
  SPEAKER_00: But the shop still needs its customer information connected to its marketing.
• 4:18
  SPEAKER_00: It needs to avoid offering a discount to someone who bought the same item at full price 10 minutes ago.
• 4:24
  SPEAKER_00: And someone needs to look after the system when a connection breaks or the information is wrong.
• 4:30
  SPEAKER_00: Our argument for Clavio is that AI could make those existing capabilities easier to use.
• 4:37
  SPEAKER_00: The company has a tool called Composer, which entered public beta around the end of June.
• 4:44
  SPEAKER_00: A marketeer can describe a campaign and Composer can help put it together using the information already available in Clavio.
• 4:53
  SPEAKER_00: This is so useful because writing an email is only part of the job.
• 4:58
  SPEAKER_00: Knowing who should receive it and what they are likely to respond to matters too.
• 5:03
  SPEAKER_00: And of course, competitors will make similar promises.
• 5:06
  SPEAKER_00: Adding AI does not give Clavio an advantage by itself.
• 5:10
  SPEAKER_00: It has to help customers sell enough additional products to justify the costs.
• 5:15
  SPEAKER_00: But what we like about Clavio in this discussion is the way its core email product is priced.
• 5:21
  SPEAKER_00: The bill depends on active customer profiles and sending allowances, rather than the number of marketers using it.
• 5:30
  SPEAKER_00: Other products have different charges, but the distinction is useful.
• 5:35
  SPEAKER_00: A business might use AI to reduce its marketing team from four people to maybe two.
• 5:42
  SPEAKER_00: If it still has the same audience and sends its campaigns through Clavio, that does not automatically cut Clavio's bill in half.
• 5:51
  SPEAKER_00: It could even make the software more appealing.
• 5:54
  SPEAKER_00: Perhaps the owner of a small business can now run campaigns that previously required an agency or a specialist.
• 6:01
  SPEAKER_00: We would expect some owners to welcome that.
• 6:05
  SPEAKER_00: The results so far give us a reason to keep looking.
• 6:08
  SPEAKER_00: In Q2 of 2026, Clavio's revenue grew 26%.
• 6:14
  SPEAKER_00: Its net revenue retention rate was 109%.
• 6:18
  SPEAKER_00: That second figure means the existing customer base was spending more overall on the company's measure, and after allowing for customers who left or reduced their spending.
• 6:30
  SPEAKER_00: Now obviously, this doesn't mean that everyone loves the product.
• 6:33
  SPEAKER_00: Some customers can expand while others leave, but it is useful evidence that Clavio has continued to grow with customers it already has.
• 6:42
  SPEAKER_00: But it would be fair to just be positive, right?
• 6:45
  SPEAKER_00: Where we become less comfortable is thinking about my own inbox.
• 6:50
  SPEAKER_00: There are brands alike whose emails I barely open.
• 6:54
  SPEAKER_00: Sometimes I bought one thing a year ago and they've been telling me about it every sale ever since.
• 7:01
  SPEAKER_00: Let's say my AI assistant starts sorting through those messages.
• 7:05
  SPEAKER_00: I will probably ask it to show me what matters.
• 7:09
  SPEAKER_00: I doubt I will ask it to preserve every promotional email.
• 7:14
  SPEAKER_00: We do not yet know how far that behavior will spread, right?
• 7:18
  SPEAKER_00: But it could change the value of email marketing altogether.
• 7:22
  SPEAKER_00: You could have Clavio helping businesses produce better campaigns, while another AI tool helps consumers ignore them.
• 7:30
  SPEAKER_00: If those campaigns produce fewer profitable purchases, the businesses sending them will eventually reconsider what they spend on.
• 7:39
  SPEAKER_00: That concerns us more than the idea that a small retailer will suddenly build its own complete marketing system.
• 7:46
  SPEAKER_00: Building software is becoming easier.
• 7:49
  SPEAKER_00: Looking after it still takes work, and most shop owners have quite enough to do already.
• 7:54
  SPEAKER_00: They do, however, have a reason to choose a cheaper provider if it does the job well enough.
• 8:07
  SPEAKER_00: Retention helps us judge that, along with their willingness to pay for additional products.
• 8:14
  SPEAKER_00: We would also watch what shareholders actually receive.
• 8:17
  SPEAKER_00: Clavio generated roughly$83 million of free cash flow in Q2, while still reporting an operating loss.
• 8:25
  SPEAKER_00: Stock compensation helps explain part of that difference, and cash is useful, but the shares issued to employees also have a cost for existing owners.
• 8:35
  SPEAKER_00: We can see a future where AI makes Clavio more useful and helps it reach more businesses.
• 8:42
  SPEAKER_00: We can also see one where marketing becomes easier to produce and harder to get anyone to notice.
• 8:48
  SPEAKER_00: HubSpot brings us back to the employees using the software.
• 8:53
  SPEAKER_00: Say you run a business with 20 salespeople.
• 8:57
  SPEAKER_00: You have spent years trying to get them to keep the customer records up to date.
• 9:02
  SPEAKER_00: Calls should be locked, follow-ups should be sent, and every deal should have a clear next step.
• 9:09
  SPEAKER_00: In practice, some of that happens on a Friday afternoon when someone tries to remember what they said on Tuesday.
• 9:16
  SPEAKER_00: An AI tool that takes care of that would be very useful.
• 9:21
  SPEAKER_00: The salesperson gets more time to talk to customers and you get better records.
• 9:27
  SPEAKER_00: You might keep all 20 people and try to sell more, or you might grow without hiring the next five.
• 9:34
  SPEAKER_00: You could also decide that a smaller team can handle the same work.
• 9:39
  SPEAKER_00: We don't need to predict mass layoffs for this to matter.
• 9:42
  SPEAKER_00: Even a business that keeps its existing team may buy fewer additional seats than it otherwise would have bought.
• 9:49
  SPEAKER_00: HubSpot has a broader business than subscriptions for salespeople.
• 9:54
  SPEAKER_00: It includes marketing, customer service, and other tools around its CRM.
• 10:00
  SPEAKER_00: But paid seats are part of its pricing, so making people more productive can complicate the economics.
• 10:09
  SPEAKER_00: What makes us reluctant to write it off is how much useful information sits inside a CRM once a business has been using it for a while.
• 10:18
  SPEAKER_00: Someone has asked for a quote, someone else has complained about a delivery, a third customer has already been promised a discount.
• 10:27
  SPEAKER_00: You want an AI assistant to know those things before it starts contacting people on your behalf.
• 10:33
  SPEAKER_00: Otherwise, you have given a very efficient tool permission to make mistakes much faster.
• 10:39
  SPEAKER_00: HubSpot already holds a lot of that information and connects it to the work people do.
• 10:45
  SPEAKER_00: That gives it a reasonable place to start, and replacing an established CRM can also be a miserable project.
• 10:52
  SPEAKER_00: You need to move the records, reconnect to other software, train the team, and hope you have not broken something that only becomes obvious a month later.
• 11:01
  SPEAKER_00: We struggle to see the owner of a medium-sized business choosing their project just because someone showed them an impressive AI demo.
• 11:10
  SPEAKER_00: But there is another way HubSpot could lose out.
• 11:13
  SPEAKER_00: The customer might keep HubSpot and buy its AI tools from someone else.
• 11:18
  SPEAKER_00: HubSpot remains the place where the rackets sit while a different company earns the money from doing more of the work.
• 11:25
  SPEAKER_00: Keeping the customer would not necessarily preserve HubSpot's share of that customer's spending.
• 11:30
  SPEAKER_00: The company is already trying to earn money from automated work.
• 11:34
  SPEAKER_00: It uses credits for tools such as customer agents with included allowances and additional credits available to buy.
• 11:42
  SPEAKER_00: The difficulty is working out how much customers will pay.
• 11:46
  SPEAKER_00: If a tool saves a company$100,000, there should be room to charge something useful.
• 11:52
  SPEAKER_00: But if the customer will compare alternatives, if several providers can produce a similar result, much of the saving may stay with the customer.
• 12:02
  SPEAKER_00: Which would be excellent for the customer, but not as good for HubSpot.
• 12:07
  SPEAKER_00: Running those tools costs money too.
• 12:10
  SPEAKER_00: Replacing subscription revenue with AI revenue may change the margin.
• 12:15
  SPEAKER_00: Now HubSpot has continued to grow.
• 12:18
  SPEAKER_00: Q2 revenue grew 20%, customer numbers grew 14% to more than 300,000.
• 12:26
  SPEAKER_00: So those figures are encouraging.
• 12:28
  SPEAKER_00: We think HubSpot has a credible chance of adapting.
• 12:31
  SPEAKER_00: We would want to see whether customers pay more over time and whether the new tools are earning enough after their costs to make it up for any pressure on seats.
• 12:40
  SPEAKER_00: Revenue growth itself is not enough on its own.
• 12:44
  SPEAKER_00: Lastly, we haven't touched on Agen.
• 12:47
  SPEAKER_00: And Achen is a different beast altogether, because they make money from payments.
• 12:52
  SPEAKER_00: If an AI assistant finds a product and buys it for you, somebody still needs to handle the transaction.
• 12:59
  SPEAKER_00: Your payment needs to be approved, fraud needs to be managed, the merchant needs to receive the money.
• 13:06
  SPEAKER_00: Now the way you find products can change considerably, while much of the work remains necessary.
• 13:13
  SPEAKER_00: That's why we think AGen is still worth considering here.
• 13:17
  SPEAKER_00: And we think that AI is actually not AD's biggest risk, but Stripe is.
• 13:23
  SPEAKER_00: If Stripe gets a customer early enough and does a good job, and we can tell you they are, why would any customers leave as it grows?
• 13:32
  SPEAKER_00: They might not.
• 13:34
  SPEAKER_00: We should be careful about the idea that successful startups eventually become large enough to move to ADGEN.
• 13:40
  SPEAKER_00: Stripe serves large businesses too.
• 13:43
  SPEAKER_00: Becoming an enterprise does not create an obligation to change payment providers, so Adgen has to offer something worth the effort of adding or switching to it.
• 13:54
  SPEAKER_00: We now have an example that supports the ability to compete with Stripe.
• 14:00
  SPEAKER_00: In the first half of 2026, AGEN named OpenAI among its new enterprise customers.
• 14:07
  SPEAKER_00: That's a very good sign.
• 14:09
  SPEAKER_00: It shows ADEN can win businesses in a category where investors were worried about its position.
• 14:16
  SPEAKER_00: But we don't know how much of OpenAI's payment business it handles.
• 14:21
  SPEAKER_00: And while this is unconfirmed, but I believe OpenAI uses both Stripe and ADN rather than relying on just one payment provider.
• 14:30
  SPEAKER_00: They use Stripe for the core subscription billing, such as ChatGPT Plus, but they added AGEN as an additional payment partner to provide international redundancy, improve authorization rates, but it is an encouraging customer win, whose financial importance still needs to become clear for ADGEN.
• 14:49
  SPEAKER_00: So, ADEN needs to win business that is attractive for ADN as well.
• 14:54
  SPEAKER_00: There is another development that connects its way to AI services and how they are sold.
• 15:01
  SPEAKER_00: Agen recently bought ARP, with the acquisition closing on July 1st.
• 15:07
  SPEAKER_00: ORP provides billing software.
• 15:10
  SPEAKER_00: It includes billing based on usage.
• 15:13
  SPEAKER_00: An AI service might charge according to how much you use it.
• 15:17
  SPEAKER_00: That creates a job before the payment even happens, because someone has to measure the usage and calculate what you owe.
• 15:25
  SPEAKER_00: ORP does that billing work.
• 15:27
  SPEAKER_00: ACEN can handle the payment.
• 15:29
  SPEAKER_00: Bringing the two together could make life easier for a business that would otherwise have to connect separate systems.
• 15:36
  SPEAKER_00: A pretty sensible reason for an acquisition, but now they have to make it work commercially.
• 15:42
  SPEAKER_00: Agen recently also launched AGen Agentic to support payments involving AI agents.
• 15:49
  SPEAKER_00: So there is clearly an effort to adapt the products.
• 15:53
  SPEAKER_00: But the part of our earlier ACEN research we keep coming back to is their ability to begin with some of a merchant's payments and earn more business over time.
• 16:04
  SPEAKER_00: If Agen improves the payment experience or the merchant's economics, the merchant has a reason to give it more volume.
• 16:11
  SPEAKER_00: It does not depend on winning the entire customer immediately.
• 16:16
  SPEAKER_00: And although ADGEN focuses heavily on enterprises and large businesses, it also reaches smaller businesses through platforms.
• 16:25
  SPEAKER_00: Its potential business is broader than only the largest companies buying directly.
• 16:31
  SPEAKER_00: Of these three, we find AGEN's relationship to AI somewhat easier to understand.
• 16:38
  SPEAKER_00: People and businesses may buy things differently, but they still need to pay.
• 16:44
  SPEAKER_00: The uncertainty is who handles those payments and how much they earn.
• 16:50
  SPEAKER_00: Stripe is a serious, serious competitor, no doubt about it.
• 16:55
  SPEAKER_00: Large merchants have a lot of negotiating power.
• 16:59
  SPEAKER_00: New ways of paying could change the economics.
• 17:03
  SPEAKER_00: Argen could benefit from AI and still be a disappointing investment, right?
• 17:09
  SPEAKER_00: That is where we end up with all three companies.
• 17:12
  SPEAKER_00: There are reasons to believe they can adapt, but we still need to work out what those businesses are worth.
• 17:20
  SPEAKER_00: A stock that falls from$100 to$30 feels cheap because we remember$100.
• 17:27
  SPEAKER_00: It's a psychological trick.
• 17:29
  SPEAKER_00: We start treating the previous price as something the company should return to.
• 17:34
  SPEAKER_00: But the company has no obligation to do that.
• 17:36
  SPEAKER_00: The stock price doesn't have to do that.
• 17:39
  SPEAKER_00: We would want to value these businesses with more conservative assumptions, slower growth, less generous assumptions than the most enthusiastic AI story suggests.
• 17:52
  SPEAKER_00: For software companies, include stock-based compensation and dilution.
• 17:57
  SPEAKER_00: We would also leave room for competitors to take some market share.
• 18:02
  SPEAKER_00: If an investment still looks attractive under those reasonable conditions, we become much more interested.
• 18:10
  SPEAKER_00: We think these three deserve further research.
• 18:13
  SPEAKER_00: We have a full Agen deep dive available on the TDI terminal for those interested.
• 18:19
  SPEAKER_00: We also have a free interview talking to employees of ADGEN and what the future holds and looks like.
• 18:25
  SPEAKER_00: I'll make sure to put the links in the description.
• 18:28
  SPEAKER_00: If you want to follow more of our research, you can create a free account at tdi-terminal.com.
• 18:36
  SPEAKER_00: That is tdi-terminal.com.
• 18:40
  SPEAKER_00: You can also find links in the show notes.
• 18:42
  SPEAKER_00: That's it for today.
• 18:44
  SPEAKER_00: Thank you so much for being here again.
• 18:46
  SPEAKER_00: And as always, stay curious, keep learning, and happy investing.