Check out the conversation on Apple, Spotify and YouTube.
Meet Linda Tong (02:02)
Aakash: Linda Tong is the CEO of Webflow. She is a product person at heart. She was chief product officer at Tapjoy, chief product officer at AppDynamics and led product at the NFL. She’s had a really cool post-product career, which I think will be an inspiration for us all. At AppDynamics, after they got acquired by Cisco, she became the general manager. Then she joined Webflow as chief operating officer, was promoted to president and COO, and now she is the CEO.
So she’s got a really aspirational career arc. She understands product and she can speak to it intimately. Today we’re going to talk about whether SaaS can survive the AI era. She has one of the most unique windows into that question, because Webflow has been around since before AI. We’re going to explore all of that today. Linda, thanks for being on the pod.
Linda: Oh my gosh, thank you for having me.
How Webflow competes with ChatGPT and Claude site builders (02:58)
Aakash: I want to start here. ChatGPT has a site builder, and they just keep talking about the stats there. Even in Claude, people are building their websites. So if you think about it, one of the first application-level companies the frontier model companies attacked is companies like Webflow. As the leader of Webflow, how do you think about that strategically, and how do you counterposition yourself?
Linda: Yeah, so we saw this coming for a long time. People started using AI to write code, and especially front-end code, years ago. What we realized was that while AI can write code, a website is more than just assembling a bunch of HTML, CSS and JavaScript. It’s actually thinking about what happens once you hit publish. That’s where all the work happens.
What we’ve been doing over the past few years is expanding the Webflow portfolio. A lot of people think of us as a website builder, but we’re actually so much more than that. The majority of our business sits after you hit publish. Everything from your analytics, experimentation, personalization and localization. But also how you think about iterating your website and actually managing it across a team of people who are using that website to drive brand awareness and growth.
So for us, how you want to build it, whether you want to use Webflow or an AI code generator or any other platform, that’s fine. That’s all great. What we want you to do is build your site, bring it to Webflow or build it in Webflow, and use this to help you manage that site. The full life cycle of the thing you’re doing to drive brand and growth.
Aakash: What is the difference between managing your site on Webflow versus inside ChatGPT sites or Claude?
Linda: Yeah, and this is actually something we talk a lot about with OpenAI and Anthropic. A lot of people create those sites there, but then what’s the management platform? What’s the UI you interact with when you want to, say, publish another blog post or update the branding? You can either just go rebuild it, where you’re just prompting it, or you can work in a platform that’s actually built to iterate. With Webflow, you get not only the AI if you want to keep prompting, but also a visual canvas where you can start to iterate and edit. As you want to push changes, you have an environment that’s built for you to actually manage those changes.
More importantly, there’s the collaboration, because very rarely is the website a one-person job. It’s usually a team of people coming in and doing a lot of things together. So for us, it’s the experience of how you collaborate with your teams to drive changes. Oftentimes we actually encourage a lot of people to use ChatGPT’s site builder and then bring the site into Webflow and iterate from there.
Aakash: Why wouldn’t they just iterate in the platform? How do you think about building offerings that make it really feel like I have to go to Webflow?
Linda: It’s actually the entire system around it. For example, when you’re iterating, you don’t have a CMS built into ChatGPT or Anthropic. If you want to add a blog post because you’ve just launched a new product and you want to announce it, you can either prompt some static page, which is not ideal, or you can come into Webflow and literally just add a blog post in a visual experience that’s built for a content writer. Just text fields and image fields. They’re editing those, visualizing it and publishing it, and they don’t have to touch the rest of the site or worry about all the things that go around publishing a change to your website.
The security you have to think about. Roles and permissions. Staging environments. These are all things that are so much harder to do in an environment where you’re essentially going zero to one constantly, as opposed to having a system that’s built around your organization and the roles and permissions of the people who have to go in and experience that digital experience. A platform that helps you manage those changes effectively.
How to carve out ground the frontier labs won’t build (06:48)
Aakash: Is there a broader way to think about it strategically? How do you carve out product areas that the frontier labs aren’t just going to eventually build themselves?
Linda: Some of it is thinking about what you can build that they won’t build themselves. But the other part is going really deep in a domain. I think the world right now is splitting. You see horizontal capabilities, and you see that a lot with the frontier labs. They are going really broad and unlocking so much for you. I am a million times more productive because of my ability to use Codex or Claude Code. But we are going vertically deep and building domain expertise.
Because with software like Webflow, you not only want to build a website. Ultimately you’re trying to drive growth or conversion or some sort of output with your site. You’re driving traffic to it, and you want signups or purchases or people to take some sort of action. The person who owns that in your business is constantly iterating it and testing it and adding to it and growing it, and they don’t necessarily know the right answer.
So you want to build a system around it that sets them up for success. That means building an entire analytics platform. You don’t get that out of the box with ChatGPT or Anthropic. An experimentation platform to test variants. A personalization platform with targeting that hooks into your CDP. Let’s say you’re running a campaign and you want to target your top 50 customers. Those capabilities aren’t built in, and that just goes deep into the domain.
So for us, we are a marketing-specific organization, and we want to support marketing teams in how they become AI-native marketing teams. I think that domain specificity is really how we differentiate in this world. Not only do we bring the domain knowledge, but everything we build starts to cater to those personas more specifically, so they don’t have to build the one-offs or the nuances themselves.
More importantly, we can take all the things we build and build continuous learning into the agent workflows we’re building, so they’re continuously improving upon themselves and getting you to better outcomes. What a user ultimately wants when they’re using AI is for it to be just as smart and just as effective, and to continuously learn. If you treat your agent as maybe an intern when you first start working with it, your hope is that it’s not an intern 60 days from now. Rather, it’s at the level of a full-time employee you would hire with years of experience. So we’re building a system that not only unlocks the capabilities, but is a training system that takes all the workflows built into it and improves them over time.
Why going vertical on marketing builds a lasting business (12:27)
Aakash: One thing you said that I want to latch on to is being a marketing-specific organization.
Linda: Yeah.
Aakash: That seems like one area where a lot of startups are seeing a lot of success. Whether it’s Harvey being very specific to legal, or different people focusing on one specific area. Is it that specificity that really allows you to create more of a lasting business? Because then you can really go deep on what the marketing workflow is, and we know the frontier model companies aren’t going to approach it that way, for one function within the company.
Linda: Totally. I think that’s a massive differentiator. If you’re horizontal, you have a much larger TAM and so many more use cases to go after. If you’re vertical, you’re really narrow, but you can actually go in and support the nuance. You can build something so specific and so catered to an audience that you get much better outcomes.
You even see this when people think about models. Your frontier lab models are great for everything, but if you want a really narrow model for a really specific use case, people are just training their own mini models because they can make them so much more effective. You don’t need a model that knows Washington was our first president. There’s a lot of knowledge you don’t have to build into a model to make it really effective for narrow use cases, and you have the freedom to do that when you verticalize.
Aakash: So were you marketing-specific before AI happened, or was that a strategic decision you made in response to seeing it?
Linda: I would say we were marketing-specific before. I don’t think we realized it, though. As a company, we always saw ourselves as a relatively horizontal product, because every company needs websites, and those websites differ in how you use them and in their complexity.
But as we were growing as a company, we found that our customers were pulling us. A lot of our evolution over, frankly, the past decade was driven by our customers. Them telling us what features they wanted, how they wanted to use the website, what they needed out of our CMS. It naturally pulled us in a direction where we realized our core audience was marketers, and specifically larger marketing teams, as opposed to solo individuals standing up a landing page.
The types of websites built on Webflow tend to be websites for larger corporations, where there are large teams of dozens to hundreds of people on the marketing team collaborating to make that website their main source of growth in a digital era. So as we started to see that evolution of how people use our product, we realized our ICP really was marketing teams.
Making the call to go all in on marketing teams (15:08)
Aakash: It can be risky. At least I perceived Webflow, let’s say five or six years ago, as serving many different functions, and most likely getting even regular users like me who just want to put a blog up. Webflow was one I strongly considered. So how do you make that decision to go all in on one specific area? It seems scary for the average founder or leader.
Linda: I mean, it’s terrifying. A big thing is recognizing that you’re not building something just for one audience and rejecting everyone else. When you’re a product person, you have to identify your ICP, your core use cases and how people use your product, and you have to build for them. Naturally you see a cascade where other people will use it, maybe not for that exact use case but for something adjacent or similar. That doesn’t mean you can’t address them. But how you prioritize what you build, and how you build, should address that audience first and foremost.
For us, it was less about spinning a wheel and picking somebody. You’re right, PMs use Webflow, developers use Webflow, designers use Webflow. What we found was that they were using Webflow for marketing use cases. Not everyone personally identifies as a marketer or a product marketer, whatever role you want, but they were using Webflow for marketing. That’s where we realized we needed to target marketing use cases and marketing teams. Oftentimes those developers were web developers for the marketing team, or PMs driving growth on the marketing team, or actual product marketers using it to update content.
Deeply understanding that about our users was so critical, and making that call was important so that we would stay focused and not get lost. Because you’re right, there are other website builders whose entire purpose is standing up really simple landing pages for individual use. Maybe a portfolio or a blog.
Where we shine, and the biggest thing that stood out for us, was collaboration. An individual can build a website. I have my own website in Webflow. But where we shine is when teams have to work together and push changes with all this complexity. We have all of that built into our platform, so they can collaborate and branch and work in staging environments, and manage this digital experience without stepping all over each other. That complexity of how teams work together and collaborate was one of the key unlocks for our customers. It forced us to really focus on where we shine, and that’s ultimately where we landed on marketing teams.
Should every SaaS company draw a box around its users? (17:46)
Aakash: I feel like there’s a really deep insight there. After looking at your own user base, you identified that a marketing team is not going to suddenly give everybody full control to mess up the security of the site by vibe coding it in Claude Code, or pull down the site, or require everybody, including the SEO guy, to learn that just to publish a simple blog post. You had this deep insight about a customer base that was pretty big for you, probably one of your business segments already, and you went really deep on them. You said, okay, this is a defensible box.
Is that the broader template most SaaS companies should be thinking about? What group of users can I draw a box around that the frontier labs aren’t really going to build product for?
Linda: I would tease that apart. There’s understanding your users and making sure you understand the shape of your business and how people use it. Then there’s picking the area of your business that makes the most sense to focus on. That’s really where we drew that box.
A consideration as we do this is whether it gives us a moat, and what our competitive threat and competitive landscape look like. That’s one consideration for us. As a vertical solution, I do think it puts us in a space where I don’t see a frontier lab directly coming after us. Maybe they will eventually. I think they all think there’s going to be one company in the future, so we’ll see.
But that’s a consideration as opposed to the goal. As a product person, you have to be bold. You shouldn’t necessarily say, if this is where everyone else is going, I’m going to go in a different direction to avoid them. Sometimes your best strategy is going and competing in that space, and I wouldn’t want to be avoidant there. It just happened that where we drew that box put us into a more specific and narrow domain that probably isn’t the first priority of a frontier lab right now.
Does SaaS survive AI? (19:48)
Aakash: So that’s what you did. If you reflect on the broader SaaS landscape, is there going to be a SaaS reckoning like the one we’re seeing in the multiples? Are certain companies not going to be able to draw that box? Are they going to be subsumed? How do you think about the broader SaaS landscape and whether it survives AI?
Linda: It’s such a loaded question. I would tease it into a couple of things. One, I think we’re seeing a bit of a thawing of the SaaS apocalypse, or SaaSmageddon, whatever you want to call it. We’re seeing that there are companies that are not at all threatened, and instead are actually going to find tailwinds with AI, as opposed to facing an extinction event because of the frontier labs.
The characteristics that make those companies sustainable, and frankly I think really strong long-term companies, start with context. People have been tossing around terms like system of record. Really what it is is that they have domain context that makes them a center of gravity. People have stored their data and their context in those systems, no one wants to take it out, and you want to continue to work with those systems.
The biggest change, and you can see this in how companies like Salesforce are operating, is whether you continue to exist as a UI solution for people or whether you just stay a system of record. Maybe the opportunity right now is software that sits on top of how people interact with you. Maybe no one wants to work with the UI they were given before in the Salesforce CRM. But the data that’s there and the structure, you don’t want to move it. And the platform built around it, with all the guardrails, the governance and the security, is great. It’s an excellently built product. No one wants to vibe code that.
So maybe the opportunity for them is making that accessible, adjusting their monetization and letting other people build UI on top of it. Maybe I’m using Claude Code to build my own workflows that talk to my CRM. I think companies that have that domain or context gravity absolutely survive.
Companies that don’t have that, where their entire value is a UI layer or an API layer that talks to other systems, that’s where you start to run into potential risk around the right strategy for your business. It’s a question of what core value someone gets out of your product. Is it differentiated? How easy is it to replicate and build? If you don’t have any real domain context in there, the speed at which someone can migrate off of you is also really, really fast. Those are the considerations I would have.
So back to your question. SaaS is such a broad category. I don’t think everything survives. But I do think the vast majority, where you’re in domain-specific areas, you have deep context and there’s a meaningful amount of data or company knowledge stored there, have an interesting moat. Then it’s a question of how you run that business to adjust for how people want to interact with you in the future.
Governance, criticality and the real cost of vibe coding your own tools (22:48)
Aakash: So we’ve figured out that if you’re very domain-specific, you can definitely create some defensibility. If you create a system of record, you can create some defensibility. Are those the two main paths? Does everyone need to do that? How do we think about the rest of SaaS? Let’s say you’re still just building a horizontal website builder. Are you doomed to fail?
Linda: The two other things I would consider are, first, governance and control. It’s probably the most boring thing. But governance, controls, roles, permissions, compliance. If there are deep governance and controls in your system and they need to be there, let me tell you how few people want to vibe code that and build their own systems for it. That gives you a bit of an advantage.
The last piece is the criticality of what you build, and whether it’s worthwhile to build internally. This is where I think the math is finally starting to come to fruition. Even if you don’t have those prior three things, it’s essentially a CFO question. What’s the total cost of ownership of paying someone to do this for you? Paying for some software that does whatever task you need, whether it’s building a website, doing some sort of accounting or managing a project management tool. All these things cost money. The question is how much you have to pay for something that meets your needs and does everything you need, versus how much it would cost you to build it.
The problem is everyone says, oh yeah, I could vibe code that in an afternoon. And the answer is, if you vibe code that in an afternoon, software, just like everything else, deteriorates over time. It’s not like you build something and it’s great and it keeps working. There are bugs and issues. As more people come on, you have to think about features and functionality, and you have to keep investing in it. So a CFO has to ask how much time and effort they have to continuously put into maintaining the software versus buying it. If the total cost of ownership doesn’t add up, no one’s going to vibe code it, because it is so much more expensive to hire great talent and have that talent build something that’s not critical to your business.
That’s why SaaS exists in the first place. A lot of things have turned into SaaS businesses, like companies that manage payments or companies that literally help you with authentication. You can absolutely build these, and people have built them on their own, but they’ve realized that managing just that part of their product makes no sense. It’s so much effort to do something someone else can build for you, when you can put your resources on things that are really specific to your business.
I think that analysis is finally happening. There’s the joy of building anything and everything on your own, and then there’s the realization that you can’t just build it, set it and forget it. You have to keep maintaining it, and the cost of doing that is so painful. So you have to do a TCO analysis, and frankly that’s where people are starting to see change.
We have a group internally we call our boomerangs. It’s folks who said, oh, I’m going to vibe code my own website builder, I’m just going to go vibe my website and do all this stuff. I’ve had a couple of CMOs call me with very colorful language telling me, “We went down this path and it’s totally effed. I need to come back to Webflow. Please don’t shut it off. We’re back.” And I’ve seen them return in about 90 days.
Is the SaaS multiple compression justified? (26:10)
Aakash: Okay. Let’s talk a bit about whether this compression is justified. Look at even the categories you just talked about, like payments. PayPal’s multiple has collapsed, and that’s one where it might actually make sense not to build it yourself. Or companies like Avalara, which figure out your tax strategy. Their multiples are collapsing.
Then we saw the two Bending Spoons acquisitions recently. We saw Airtable go from an $11 billion valuation at its peak to a $1 billion acquisition. We saw Miro go from $17 billion to $1.3 billion. That’s like 100x multiple compression, or something like that. So is that justified?
Linda: I think it’s partially justified. The way people do the math is, if your growth is starting to slow, and let’s say before you were growing 50%, you could get a 10x multiple, because they figure that to actually reach that valuation you only need four or five years. Now if you’re growing 10 or 20%, you need way more time to justify the multiple. So you now see 2 to 3x multiples for them. The math itself isn’t wrong.
The thing worth noting is, one, a lot of these multiples that are compressing are coming down from an era where people were giving 30, 40, 50x multiples. As a market, we have booms and busts, and we’re coming from a time where things were frankly overvalued. So we’re seeing the compression from that.
Two, we’re seeing fear. If you look at the public markets, which actually lead the private valuations, they’re starting to thaw. Before, you saw major SaaS companies like Salesforce and even Adobe down significantly year-over-year. Now you’re starting to see their stock prices recover a little. They’re not fully back, but you’re starting to see some recovery, so people are realizing there’s more sustainability in these businesses. But with these very low 2x, 3x multiples right now, I do think there’s still a hangover of the fear, and I do think there’s upside.
Frankly, PE is probably the most, I would say, vicious of all the financial methodologies. They want to buy low and sell high. The entire PE model is to buy as low as possible, and you would only buy if you believe that in a couple of years you can flip it for a 2x, 3x return. So clearly someone is seeing a lot of value in these products. They have great brand recognition, they have great products, they’re still growing, and they have users who have embedded their businesses in them. The buyers believe these things aren’t going to decline, but rather that they can hold them steady, create operating profit and turn them in the coming years. So I think even those acquisitions are a sign that one of the most prudent markets is now acquiring these companies, because they see opportunity on the horizon.
What Salesforce got right by going headless (29:24)
Aakash: Let’s take those two examples a little deeper. Salesforce and Adobe. What are they doing right that’s starting to allow them to recover? For example, Salesforce went for this headless model on top of their system of record. Is that a good move for companies to take?
Linda: Absolutely. When I saw them make the announcement, my first thought was, they nailed it. It’s so important. In my business alone, we spend a ton of time putting so much data into Salesforce, and the painful thing is people going, ugh, I have to log in, I have to go find it.
The problem with software is it was built for what I would call the static era. Probably the easiest analogy, and I’ve been telling people about this, is maps. Two decades ago, if you wanted to go from here to, let’s say, the Moscone Center in San Francisco, you would print out a map on MapQuest. It would give you turn-by-turn directions, and you’d follow those directions to a T. If you missed one, sucks for you. You’re lost. Nowadays you just get in a car, use your phone or GPS or whatever, put in your address, and as you twist and turn, it constantly updates for you. It’s iterating, and it’s personalized to your journey.
Software needs to take that exact same evolution. What’s been unlocked now with AI is this concept people are talking about called malleable software. I think of it as hyperpersonalized software. Right now, when you and I log into Salesforce, we’re going to see relatively the same thing. The way we navigate is the same. We have the same headings, the same structure, and search returns the same results. There’s some personalization in terms of access controls, but the UI is the same. That’s ridiculous, because there are so many features and so much functionality in Salesforce. It’s like MapQuest. It’s the same static thing we’re interacting with.
What if, suddenly, and I think headless now opens this up, the way I interact with software can be a lot more malleable for me? I could access it via an MCP and say, hey, I want to do this, I want to do that. It’s not showing me all of the UI. It’s surfacing only the things that are relevant for me, and it’s navigating me through the different features and functionality of Salesforce more effectively than anything else.
When you have dense products like Salesforce, with thousands and thousands of features built over decades, as a user you don’t want to have to know every single feature, where it’s buried and how to find it. An MCP server can easily understand that, index it, navigate you through it effectively and make the user experience infinitely better. All of us need to take a page out of that book. As we build these feature-rich products, how can we create user-centric experiences that don’t force us into the age of MapQuest versus modern-day Google Maps?
How to avoid becoming a database with rules (32:12)
Aakash: How do you avoid just becoming a database with rules, if people are just accessing you with MCP and the Claude UI?
Linda: It’s the context. We actually just posted about this today. Internally, our team does a lot with how we use our own data to drive decisions, both on the features we build and on how we run the business. Every team has said, oh, I can use AI to write SQL queries and query databases. But it’s not as simple as saying, hey, you’re a database, and now people can query it. You want to put context around it, so it’s smart, it’s intelligent and it’s constantly learning.
Software needs to be the same. You need to build not only your harness around it, but also a context engine for your users. As they interact with it, you can start to make recommendations. You can enhance the feedback they get. You can change their behavior over time. So instead of it feeling like a thing you’re just querying, this massive hoarder’s yard of all your data that someone is seeking information out of, it actually becomes a foundation on which you can build agentic partnerships.
I imagine a world where, when I look at my org chart, my team will be made up of my standard, typical employees, but also a team of agents that work with us. Those agents will have context, will have learned and grown, and will interact with us while we work on top of those foundational products and data stores, continuously improving and allowing us to run our business more effectively.
Webflow’s $4 billion valuation and the multiple collapse (33:50)
Aakash: You’re the CEO, so I can ask you these tough questions. You raised at a $4 billion valuation at that same time as the $17 billion Miro and the $11 billion Airtable. What has happened to your valuation since then? Your growth? What can you tell us from the inside, as an operator, about how you’re experiencing the SaaS multiple collapse?
Linda: Yeah. Luckily, and this is kudos to the founders, because I’ve only been at the company for four and a half years. When we raised, it’s always fun to get a big valuation, but more importantly, the way they ran the company was really efficient. Operationally, we have not needed more money. We’ve been relatively profitable, then we did a little bit of burning, but we still have plenty of cash in the bank. The business has been very close to break-even for quite some time, and we’ve been signaling free cash flow. So it’s a really healthy business. As a result, we haven’t had to go raise more money or do any sort of down round or revaluation.
However, there is the reality that valuations from 2021 and 2022 just aren’t believable. So we have an opportunity to grow into that valuation, and that’s what we’ve been focused on as a business. We’ve been thinking about how we strategically pivot and iterate and innovate in this time when chaos is reigning in the SaaS ecosystem. A valuation is a point in time. What really matters is whether you’re building a long-term sustainable company that can continue to deliver value and continue to grow.
We have looked at areas of opportunity and growth and leaned into them. Over the past few years, one of the biggest shifts we’ve made came from customers pulling us upmarket, and that’s the best, when your customers tell you what to do. We had customers saying, hey, we joined Webflow when we were just coming out of YC. We’ve now scaled. We need these features and functionality. We need better security. We need governance. Do you have SOC 2? Those customers were pulling us because they didn’t want to move to one of the other, older platforms. They were like, please God, don’t make me move off of Webflow. So we grew with them, we started moving upmarket, and we’ve made a really big enterprise push over the past few years. That’s really paid off, and we’ve seen accelerated growth there.
The other thing we did was see how the market was moving back in 2021 and 2022, and we made a big shift to become a multi-product company. Instead of just being a web builder, we acquired a couple of companies, one called Intellimize and another called Vidoso, and we added products. So we can not only build a website, but offer an analytics platform, an optimization and personalization platform, and an assets platform that handles media, video, imagery, editing and asset management. We’ve also built a lot internally, rebuilding our foundation to be what I would consider AI-native and AI-ready.
And we’ve been innovating. We just had our big user conference a couple of weeks ago, and we pre-announced a product that’s coming real soon. We have a handful of people in our research preview. Source is our next innovation. It’s an orchestration and collaboration layer that sits on top of all of our products and lets people get custom views for how they want to work inside of Webflow. It manages not only everything that has to do with the web, but frankly, it’s expanding to become everything in marketing. It’s an agentic marketing platform that will run those workflows and build agents and views for our customers, to be their partners in running and managing their marketing teams.
What an agentic marketing platform actually does (37:36)
Aakash: Can you say more? What does it mean to be an agentic platform? It sounds like this is a really big part of the bet you’re making, going back to whether SaaS can survive the AI era. How can people take lessons from the positioning you’ve chosen?
Linda: There are two things within the way people use Webflow today. There’s building your site, but the majority of the time is spent once you’ve built it. It’s the tweaking, the edits, the changes. Let’s use a really simple example. I’m launching a product next week. I need to put together a blog post. I need to create the assets for it. I need to get approvals. I need to make sure the tone and voice of the content is in line with my brand. I have to make sure the features are right. I probably want to load a demo on there, so I have to work with the product team to build it.
That is a lot of work that happens in a mix of Google Docs and Jira and Slack and everything else. Ultimately, some person comes into Webflow, loads that content, pushes it to a staging environment, gets approval and publishes it. You’re timing it for the moment of your launch, because maybe it’s 9:00 a.m. or 6:00 a.m. Pacific on the day of the product launch, and then you’re kicking off a bunch of social efforts.
We saw pieces of that workflow in Webflow, and we wanted to make the parts that touch Webflow really easy. And they are. You can load content in there really easily. There are views catered to the content writer. There are approval workflows. But we realized so much more of that work happens outside of Webflow, and frankly it would be easier if it happened in one place, versus some poor soul coordinating the whole marketing team through this effort. That’s what we think of when we talk about an agentic marketing platform.
We imagine a world where a user or team can come into Source and kick off that work. They could say, okay, we’re launching this product, we need to plan a product launch. You can start to put together all the different assets you need. What’s the bill of materials? Marketers are always talking about the BOM. What’s your BOM? I’m going to need these assets for social. I’ll need this content for my blog post. I’m going to need this content for a press release that I’ll share with some PR firm under embargo.
They can create all these assets in Source, and it’ll use everything we know about them. Going back to the domain intelligence I talked about before, we have their brand operating system, their tone and voice and their design guidelines. Those things will run as agents that look at all those materials and assets and generate them in line with your brand. Our asset manager can create variants of whatever social images you have, or a video clip you want to edit. It can generate it for you and store it in one place. Then people can come in, edit, comment, provide feedback and iterate. When you want to publish, we can hook into all those systems and publish on your behalf.
All the collaboration that was hidden across these different channels can now happen in one place. It’ll be a mix of people coming in and doing the work, and agents and products we’ve built working together to move this through those workflows. That asset agent could be one where you say, hey, I want to use this baseline video and create assets for these different channels. Our asset agent could cut up that video and create an Instagram Reel, a TikTok video, a LinkedIn video and one for YouTube for your blog. All of that can be done agentically.
That’s what I mean when I talk about an agentic marketing platform. It’s a team of agents that pairs with your team to execute a lot of the work you already do, but does it with the intelligence of your business, with context, and in one place so your team can manage it more effectively.
Aakash: So it feels like you’re productizing what an AI-native founder might do, so that a big marketing team in an enterprise can actually do it.
Linda: Absolutely.
Did Webflow’s growth slow down? (41:28)
Aakash: Let’s return to that tough topic, because you’re the only one who can talk about it. Did Webflow experience a decrease in growth? Look at Miro. Like you were saying, they were actually operating really efficiently, basically profitably, is what we learned. But what really killed them was the growth rate decline, from something like 100% to 10%. Did you experience that as well?
Linda: For sure. And I think it’s twofold. If you think about the timeline, a lot of companies had COVID tailwinds. You’re coming out of COVID and your growth is accelerating, and then it starts to decelerate coming out of COVID, because your ability to keep up that rate of growth, and frankly that acceleration, gets harder as your book of business gets bigger. There’s natural gravity that pulls you down. You should expect some slowing growth as you hit scale.
The other thing we saw was a slowdown as we shifted our market, because a lot of our business was coming in via our self-serve go-to-market motion.
Aakash: Yeah, your PLG motion was legendary. Now it’s all enterprise, from what I’m hearing.
Linda: Yeah, we’re seeing so much growth from enterprise. It was a big shift in how we’re actually driving growth in our business. We started to see many more customers coming in at much higher ACVs. We’re seeing acceleration on the sales side, but relatively flat to continuous growth on our self-serve side. So blended, you start to see some decline in your overall growth rate.
What’s been great is that in the areas of innovation, we’re seeing that start to prop it back up and drive acceleration in our business. You think about roller coasters as either going up or going down. Instead, what we were seeing across the business was all these ups and downs, all these bumps. When you average it out, we were continuing to grow. The question is which areas are driving acceleration and which areas are creating friction.
Aakash: And when you shift to operating a business more profitably, that’s also going to be a gravity drag, right?
Linda: Absolutely.
Building a default alive company instead of raising again (43:33)
Aakash: You have to make this decision, I guess, with the co-founders and founders. Are you trying to go out and raise again at some point once you grow into your valuation? Or do you want to grow more profitably and sustainably? Because once you raise at a $4 billion valuation, people think, oh, they’re going for the IPO, they’re going for the big outcome. Do you have to make that strategic shift with the board and your founders? How have you managed that personally?
Linda: For me, and this is just old school, building a sustainable, default alive company is how you should always build a company. We have plenty of cash. It’s not an operating problem. The only reason I would want to raise more is if we had a really specific reason for that cash. You raise because maybe you want to do a massive acquisition, or maybe you want to make a massive investment and go big in an area. Some people raise because they can. That’s just not my MO.
As long as we can build a sustainable, high-value company for our customers and continue to drive growth, that’s how we’re going to run it. Then the outcome, whether you get acquired or IPO, those are moments in time. My hope is that we can stick to our guns on the original mission of Webflow. Right now we’ve been talking about building for the teams that are shaping what’s next. More importantly, I was just mentioning the Google Maps analogy, and we believe the future of the web is a living web. I want to position us in a way that we can unlock that vision. Whether that’s joining forces with another company or going it alone and finding a path to IPO, it doesn’t really matter as long as we can continue to pursue our vision.
How Webflow’s product team builds in the AI era (45:16)
Aakash: We’ve been talking a lot about product strategy, but now I want to really triple-click into the product side of everything. We’ve identified your product strategy and how you’ve shifted it, whether that’s being more specific on marketing, building an agentic platform or making it available headless. So we’ve covered the product strategy part. Specifically, on how you’re building product, can that also be an advantage? Can you embrace the new ways of building product, whether that’s AI prototyping or AI evals, to grow into this valuation faster?
Linda: For sure. I am so impressed with our product team. I’ve got this amazing leader, Ben Haefele who blows my mind every time I meet with him. I’ve been out of the day-to-day practice of being a product manager for some time now, so it’s been really fun to get visibility into how our team is transforming and to help be a part of shaping that change. But he’s really the one who’s hands-on in this process.
I would say our product team is unlike any product team I worked with decades ago, when I was building myself back in the early days. PMs would talk to customers, write a spec, review the spec, have meetings and talk through these things. Then designers would interpret it and create some prototype. Engineers would go through the clickable prototype and build essentially their first working prototype, and you’d iterate from there. As much as we hated the concept of waterfall, it was very waterfall.
Nowadays our PMs start with a PR. They’re literally coding and building prototypes. The way teams communicate is so much more of a swarm as opposed to a waterfall. PMs, designers and engineers are building together, solving problems, testing these workable, codified prototypes, and then iterating on them, formalizing them and productizing them. The way they build is so different. Whatever you want to call it, AI-native or AI-built, AI has completely changed the way everybody in our company works. I think it’s made us more efficient and effective, but more importantly, it’s allowed us to unlock a very different level of speed than we’ve ever seen before. That’s allowing us to race headfirst into the future and unlock a much bigger vision for our company.
Measuring the results of an AI-native product team (47:42)
Aakash: This podcast has grown basically on the back of teaching PMs how to use these AI tools. You sit on the other side of it, though. We had Rachel, your former CPO, on one of our most popular episodes ever. So she helped enable the product team for that.
Linda: Absolutely.
Aakash: But when you look at the results, are you seeing results from that? Are they moving fast? Can you quantify it? And if I’m the PM doing that, how do I make sure we achieve the same results Webflow’s product team achieved?
Linda: We are absolutely seeing the results. The team is moving faster. The ways you can measure it aren’t perfect. Product is so nuanced, so I hesitate to say you should just look at PRs per engineer to see how much you’re shipping, or at how many more features you’re launching. I’ll give you the immediate flip side. I might say, oh, I shipped 100 features this quarter versus 25 last quarter. And then you’re like, yeah, but were 80 of them stupid? So there is a quality question.
Where I’m seeing the team go really fast is, yes, our shipping speed has improved significantly, and I’ve seen that. But the other thing that’s really great is this. People talk about AI in terms of velocity, just going faster, and oftentimes it equates to quantity. Shipping more things, doing more things, writing more code. The answer isn’t necessarily that more is better. It’s that more with intention is better. You want to intentionally ship the things you know are valuable.
What our team does really well is spend more time with our customers than they ever have before. So the things they’re shipping are landing, and the things they’re building are informed by our customers’ needs and wants. If you take a pie chart of their day and break it up, I’m so happy to see PMs so much more in the field and talking to customers. We have a group of what I call our early testers, or insiders, who are Webflow experts. They give us direct feedback in our Slack channels, and we expose everything we’re testing to them to get their feedback. Our team spends so much more time talking to them, iterating and building, and they can build so quickly that they get quick feedback. That’s what has really changed our product team, and frankly our entire EPD organization.
At our conference, Webflow Conf, just a couple of weeks ago, for the first time we had two keynotes, because we just couldn’t fit everything in one. We had our main keynote, and then a builder keynote where we talked about everything we were shipping. We talked about 140 new quality-of-life improvements, which is massive, and they’re the top things our customers have been voting on in our wish list.
The other thing we did was announce what we called our Paper Cuts Bar. We flew a bunch of folks out and had a bunch of our teams on call, and customers, anyone at the conference, could go to this Paper Cuts Bar and say, hey, I want this feature, or, this is super annoying for me inside of Webflow. They could submit all these asks, and our team would sit there and fix them on the spot. It was such a hit, and we got so much feedback and engagement with our customers. More importantly, it was awesome because our whole team was able to jump in and build. When everyone is a builder and they have customer context to build with, the productive value you’re able to generate out of it is magical.
Has shipping faster translated into business outcomes? (51:12)
Aakash: I’ve been preaching this, and of course we’ve had lots of product leaders on the podcast. But where the rubber ultimately meets the road is, has it accelerated growth for you? Has it improved the type of investor conversations you’re having? Has what the product team, or really the EPD org as a whole, is doing by shipping so much more translated into business outcomes?
Linda: Definitely. A good example is that we launched our MCP server last year, and it is unheard of to launch a product and see such penetration. The growth and customer adoption has been massive. The first thing we measure is the number of customers using our MCP server, which has been growing consistently and sort of exponentially. The second is how it changes behavior in our product. An MCP is just one way to access a product and do the things you always did. It’s a different user experience.
What we’re seeing is an increase in what we call our publish rate. It’s how often someone comes in, does a series of activities and then publishes or updates their site. We’re seeing people publish at a higher rate as a result, because they come in, make changes and publish. That’s a really strong signal, because we’re seeing deeper engagement with our existing customers. And new customers coming in are getting to publish much faster. Those changes show me that the products we’re building are stickier and better, are driving better engagement of our existing base, and are driving faster acquisition of new customers.
How AI changed the PM job (52:43)
Aakash: Those sound to me like OKRs individual PMs might own. It sounds like they’re able to move their OKRs more quickly, and exceed them more quickly, because they’re embracing this leaner way of working. I remember when I started as a PM, and this is just the truth, when I was a senior PM I probably spent 75% of my time writing PRDs, writing status updates and responding to emails. But now with AI, like you’re saying, you can be more customer-centric than before. You might be able to spend 50% of your time with customers, which I never did in my whole career. So it’s amazing.
Linda: It is a complete change. My first go as a PM was actually at Google. I did it as a 20% time project. I was working on two projects. One was a revenue tracking system at Google and the second was sports. If you ever searched for sports scores in Google and saw something like, the Bills are playing tonight, if you searched for the Bills schedule, you’d see the schedule and the time of the game. If you typed in scores during the game, you’d actually see the live scoreboard. That was something I built. It was a sports OneBox.
I remember being a PM and thinking the job was literally constantly writing a spec, scheduling meetings, getting people to talk, taking notes in the meetings and sending them back out saying, these are the action items, this is what we’re going to do, here’s when we’re shipping. Reviewing a blog post for when we announced it. It was important, because you were the glue for how teams came together to execute a vision. But it was so much documentation. It was so much connecting dots.
Now with AI, you can build in constraints that take your core thesis and your purpose and build them into how you work. So you’re not constantly having to say, remember, these are the OKRs, these are the goals, these are the metrics, are we all aligned here? AI can actually be your backstop for that. So you’re just building, and you can say, make sure everything I’m thinking about here falls in line with those goals. It’s amazing to see how the workflows have changed as a result. And there’s no surprise that suddenly PMs have more time to spend with customers and validate these things.
Honestly, I’m so jealous. I wish we’d had this AI moment back when I was a PM, because I’m so jealous of how you get to work and how you get to build. I was one of those PMs who wasn’t an engineer, so I had to self-learn. The idea of being technically dangerous but not good enough to be an engineer is something I think will be really perfect nowadays, because now I can actually start to build prototypes. That would have been magical two decades ago. Wow, I’m aging myself.
How funding product teams has changed (55:29)
Aakash: You have to make decisions around funding teams and funding product areas, and thinking about the composition of the product areas you fund. Traditionally, it was something like five to 10 engineers, a product manager, and maybe a fractional or full designer. Is that still how you think about funding a product area? Has that changed?
Linda: It’s totally changed. Teams are much, much smaller. I think this is more akin to how people thought about building teams when microservices got really, really big. When Netflix first started talking about how they rearchitected their whole system into a microservices architecture, you had these tiny teams that were isolated and mutually exclusive, but had points of overlap. The way you need to design your org is to break down what you need to do into independent pieces and then understand where those pieces come together. You can have much smaller teams that run independently in those spaces.
Honestly, it’s more like two, maybe three engineers, a designer and a PM. You have these smaller teams. When you say a pizza team, this is legitimately maybe one pizza, or if you’re on a GLP-1, half a pizza. But you have all these pizza teams that can now run independently, and then you need an architectural team that really has oversight, understands where these things come together and helps you plan what the independent pieces are.
It’s smaller teams going faster than ever. As a PM, you’re always thinking about four things. Time, scope, quality and, ultimately, resources. You have to calculate time and scope fundamentally differently, because your resources are so much more productive than they’ve ever been before. So you’re not sacrificing time and scope. You’re saying, actually, I can go just as fast and do just as much with a much smaller team. So you’re really thinking about whether you have the right people, and the best people, who understand what they need to do. And quality should be a non-negotiable. Oftentimes it isn’t even a consideration. As long as you say, we’re going to have this level of quality and we have this great team, your time and scope end up being quite elastic.
Every PM is an AI PM now (57:53)
Aakash: Does the profile of who’s on these teams change, specifically from a PM point of view? There was this title that was really big in 2024 and 2025 in the PM field, AI PM. Are you looking for AI PMs now?
Linda: No. I think we missed that hype cycle. It got really big, and then I think everyone realized that every PM has to be an AI PM. In the same way, every job description in our company now has requirements around how AI-pilled you are. How are you using AI to do your job? It’s become such a non-negotiable that it would be redundant to put AI in front of any title. The expectation is that you’re going to use AI to do your job.
I think it’s become so common now. It’s the same way that you don’t even see job descriptions anymore where people say you have to be able to type a minimum number of words per minute. That used to be a thing. You have to be able to type 200 words per minute. You don’t even put that in a job description anymore because it’s a baseline expectation, and that’s just how it is for us as well.
So when I think about PM capabilities, of course I’m expecting people who are curious and able to use AI. But more importantly, AI itself is changing. Curiosity and a growth mindset are the two most important things, because I want people who can learn as technology evolves. I want people who are pushing the needle on what we can do and how we can use technology to do it better.
How to stay AI fluent without spending all day on it (59:20)
Aakash: The one difference with typing is you only have to learn typing once. With AI, it feels like a full-time job just to stay on top of it. Oh, Astra is now the best model, I guess I need to switch from Claude Code to Codex. It’s exhausting.
Linda: It’s totally exhausting.
Aakash: So how do people manage that? If you’re just a regular senior PM watching, how much time should you be taking to learn these new tools? Should you be learning AI tools your company hasn’t approved yet, off in your personal time? What’s the right way to think about staying AI fluent if I wanted to land a PM job at Webflow?
Linda: Yeah, honestly, even I struggle with this all the time. It’s a question of how much you should be investing in your own self-improvement across the board, especially in an environment where something is launching every day. I first try to stay abreast of this stuff by basically subscribing to every newsletter. I wake up, I look at Twitter, I read these newsletters, I look at what’s changed, and then I dedicate time to go play with these things. You could spend your entire day trying to stay relevant, and the very next morning you’re irrelevant again. So I think it’s about finding that balance.
Technology is going to continue to change. The pace at which it’s changing is exciting and terrifying at the same time. But I think this is an age-old problem of staying on top of stuff. Before, it was easier because innovation happened in longer cycles, so you could take time to learn things. Now it’s happening in shorter cycles. You have to balance self-improvement with a focus on your area of specialty. Otherwise, every person who could potentially become an employee at Webflow could spend their entire day learning anything and everything about AI, but maybe not be as relevant to their job.
You still have to be a good PM. AI is just a component of it. It’s a tool set for you. You still have to have great judgment. You still have to have customer orientation. You still have to be able to think strategically and then translate that into structured design and thought processes. You have to collaborate well with others.
So if you want to be a PM at Webflow, or frankly a PM anywhere, one, you need to make sure you’re investing to be great as a product manager. Two, I think you have to invest to be great at the domain you want to go into. It helps if you really understand the business you want to go into. And last, keeping up on the skills. I think it’s awesome when companies give you time and resources to do it on their dime, but every person should have some amount of internal motivation to go build these skills themselves.
I encourage people to experiment and test things. I don’t expect you to be the foremost expert on everything, but I do expect you to stay abreast of the times. Maybe you don’t know all the edges around Astra yet, but you should be familiar with who the big players are, what the different models are and the core areas they’re better at, and keep abreast of how it’s changing. Then, where things are most relevant to you, that’s where I would dive deep and say, actually, this is an area I should go deeper on. For a PM at Webflow, they should go deeper on things like WebMCP and the standards that are coming out on how agents will navigate websites, because that’s highly relevant to our space. Studying models that are improving music generation is maybe less relevant, because that’s not our space, but it’s interesting.
How Webflow interviews PM candidates now (1:02:48)
Aakash: The last area I want to touch on. I think Ben and Rachel are probably closer to the individual PM hiring, but you would certainly be close to the product leader hiring. How do you evaluate prospective candidates at Webflow for these skills? Has your interview process changed? Has what you’re asking in interviews changed?
Linda: What we ask has changed, for sure. We definitely ask people what they’ve built. The best way to test things like curiosity and a growth mindset is to ask people what they do outside of work. When they have free time, where do they spend it? Are they learning something new? Do they have a passion? Are they pushing themselves to learn and grow? We ask a lot about that. Oftentimes we ask, what’s something you’ve built with AI, or what’s some way you’re using AI to improve your life? That gives us a lot of insight into their own personal drivers, how curious they are, their technical capability and where they experiment on their own.
The other thing we’ve always done at Webflow is the one-on-one interviews, where we have certain criteria we look at. AI fluency is one area, domain competency is another, culture fit is another. We look at all of those in the individual interviews. But we’ve also always done a final-round working session, where we give someone a prompt that’s relevant to the role they’re applying for, and we have them go through a problem set with us in a group setting. It might be a group Zoom, or they might come in person.
That’s a really, really important part of the interview, and we’ve kept it since the beginning. I remember when I first interviewed, I had to do it, and I thought, this is weird. But the number of people we’ve rejected after that interview has been quite high. At the end of the day, you’re bringing people together to be on a team and work together. You have to be able to see, can I work with this person? Can I work through problems with them? Can we solve things together? Or is that going to be terrible? No matter how talented an individual is, you’re hiring them to be part of a team, and if they can’t work well in a team setting, that says a lot.
That’s been one of the critical things. I’m really proud of the people we hire at Webflow. I think we have some of the best talent out there, and part of it is because they’re not only individually quite amazing, but they also work really well together as a team. It shows in how we execute as a company.
Wrapping up (1:05:13)
Aakash: Well, guys, we’ve covered their product strategy. We even touched a bit on their corporate strategy. We’ve covered what they’re looking for in PMs. I hope you got as much out of this episode as I did. Let me know in the comments below. Should I have more CEOs who were former product people on, or should we stick to the product audience? We will see you in the next episode. Linda, thank you so much.
Linda: Thank you so much. This is awesome.