Design as moat, AI marketing infrastructure, and why AI design tools aren’t ready

We started as the SEO for AI search company, building an analytics layer for brands to understand what people are searching for, what the outputs are, and how they’re actually showing up. This was timed well, and we built it at a high quality level, fast. That’s led to a lot of our success.

But then we learned… once you know how you rank, the next step is taking action. You always want to improve. The way to do that is writing the most effective content and shaping your digital presence to show up well in ChatGPT and other answer engines. LLMs opened our business in two ways, as a consumer product now approaching a billion monthly active users, and as a technology we could apply directly to our industry.

The main thesis of Profound 2.0… how do you help marketers work more efficiently and harness AI to do their work? We want to be the AI-native application layer for marketers, the same way developers have Cursor, lawyers have Harvey, finance has Legora. Profound is that for marketing.

We don’t love comping directly to SEO, but for this example… in 2024, $93 billion were spent on SEO. Only $3 billion was analytics, tools like SEMrush, Ahrefs, Brightedge. The other $90 billion is services. Hiring an SEO agency, a digital marketing agency to do research, write content, build your website. That 3-to-90 is a 30x figure.

If we succeed in bringing AI agents to help marketers do their work, that’s an enormous market, and we’re well positioned to go after it. We already work with Uber, Apple, Walmart, Shopify, Ramp, Brex, Airbnb, Meta, Amazon, and a lot of other enterprise brands. Now it’s about upselling the agents product to the same customer base.

We started quite opinionated. “Here are the strategies you should follow.” That’s shifted. As our vision got broader, we want to be almost completely unopinionated. Strategies are always changing. The best ones people are going to discover themselves. We want to give marketers the creative liberties to find those strategies and then execute on them with Profound.

Profound Agents is essentially an agent builder. Deterministic workflows, autonomous agents, full integration with your Profound data so you know exactly how you rank on any given topic, and deep integrations into all your marketing tools. That last piece is where it starts to truly differentiate.

Eventually, yes. In about three years, we won’t be doing just web text, we’ll move toward image and video generation as well. Right now we have analytics and content generation. The next thing we’re building is a content management system. Then you own the ability to publish. I joke that we want to own the means of distribution.

So the full stack becomes… analytics, generation, publishing. Then you expand the content types, images, video. And you expand across marketing functions. Product marketing, performance, social, brand. It becomes a very different company in three years.

Four things company-wide. First, you have to be an owner. There’s not much management at Profound. You carve out your area, set your own roadmap, and are responsible for what happens there. That attracts people who want real impact. Second, curiosity. We’re building things that have never been done before. You can’t repeat a playbook. You have to want to go learn.

Third, being a really good person. We’re quite strict on not bringing in the 100x engineer who destroys culture. A company, if you boil it down, is just a group of people in a room. You need to be excited to collaborate with your teammates, to be friends with everyone, to be kind. People come in and say it’s the most welcoming place they’ve ever experienced, that’s not an accident. Fourth, you have to work hard. We’re quite infamous for Saturday posts.

Three things. Visual skills. You need to know how to make things look good. That’s the taste aspect. Second, product design… do you know how to design for the business, not the user? That’s a critical distinction most designers lack unless they’ve worked at Meta, Uber, Coinbase, places that actually drill it in. Third, design engineering.

At Profound, you need two of three. I wanted all three at first. I had all three, so that was my bar. But I woke up. Those people exist, but they have the freedom to go anywhere and you’d have to pay insane amounts. Two of three is the right bar, and we make sure we’re not over-indexed on any one combination.

We have rough ratios in mind, but the real philosophy is… if you find someone good, hire them and figure it out. When you’re this picky, good at the job, the four soft qualities, and the role-specific criteria, you’ve narrowed to a very specific set of people in the world. It’s so rare when you find one that it’s almost always worth it. They’ll figure out where they fit.

Figma, of course. But a lot of our designers are getting deep with Codex and Claude Code, both in standalone prototype sandboxes and directly in our production app. We’re a feature flag shop. No staging environment. Everything is a Boolean toggle. Should we show this, should we show that. Designers can start shipping directly within the app, and it takes about 10% of the time it used to, now that you have AI tools.

One other example. Our subway ads. All the imagery is AI-generated, done with Higgs Field. People ask us how we got those photos done. We didn’t. If you look closely, one of the images shows a subway map of New York in the background, it’s complete nonsense. Nobody caught it.

We’re in a whole new industry. People are skeptical, anxious, wary of vaporware. If you build a beautiful front-end brand and product that delivers a premium experience, that squashes a lot of the preconceptions people have, and it tremendously helps with sales. You still have to do everything else well, product, engineering, sales. But the design layer is a force multiplier.

It’s also a protected advantage. It’s very hard for a founder with an engineering background to just decide “we’re going to have great design.” How do you actually do that? You might find a good designer, but that designer thinks… this person doesn’t really care about this. And they’re right. It’s hard to execute without genuine conviction at the top. So if you have that advantage, lean into it hard.

Yes. I’ve learned that everything has to come from the founding team, or someone very close to it. It’s all about alignment of priorities. You can’t retrofit it.

Controversial opinion. AI design tools are so many years away from being useful. I’ve never found them to produce good outputs, or an interface that makes me move faster. I’ve spent serious time trying.

When I spent two years looking for a business to build before founding Profound, the first lizard-brain idea as a designer is always “build a design tool.” I looked hard at it and realized… there’s no way this is going to be successful, or if it is, it’s a six-year grind. I was not interested in that.

There’s so much potential GDP being wasted because brilliant people love design and think they’re going to follow their passion and build a design tool. And it’s just not going to work out. I think it’s almost a whole other topic, smart entrepreneurs living in industries with physical limitations to their success, wasting their time. That’s what I realized.

I like ChatGPT for asking questions. But I’m never going to open a chatbot and say “order me an Uber to 248 North 8th Street.” That’s just stupid. And I think that applies to so many categories. Someone has to figure out the interface problem, it’s just not clear yet who or when. In a different world three years ago I would have loved to go figure that out. But it’s going to be such a grind with no results at the current timing. I’ll let someone else figure it out.