Why You Should Disrupt Categories, Not Create Them

Founders have long aspired to create products so innovative that they spawn entirely new business categories. While there’s allure to such dreams, it can be more strategic to build a company by disrupting existing categories. That’s the argument from James Evans, founder of Command AI (which was recently acquired), who outlines his framework in a four-step playbook in this essay. It’s the first for Every’s new Thesis column, which will bring thought-provoking big ideas every Monday to start off each week. Evans’s advice might feel counterintuitive to traditional markers of success in tech, but he’s found that such thinking can save founders years of hardship and deliver results sooner. It’s scary, but sometimes it’s better to lean into competition.—Kate Lee 

As a founder, one of the strongest 180-degree turns I’ve made is my belief that, with rare exceptions, trying to create an entire business category from scratch is a much more arduous, wasteful, and potentially disastrous path than trying to disrupt an existing one from the start. 

Category-creating businesses like Nvidia, which pioneered graphics processors, can become massive, world-changing operations—Nvidia’s GPUs are powering the AI revolution. But the odds of starting your business with paradigm-shifting aspirations and succeeding are lower than most founders realize.

It’s a lesson I had to learn myself. Four years ago, my company, now called Command AI, started as a product called CommandBar that plugged into other software and let users summon a magical search bar with cmd + k or ctrl + k. (Command AI was acquired by the data analytics company Amplitude earlier this month.) 

Users could then search the software using natural language. We felt that most software products were—and still are—too hard to use: There’s too much clicking around, too many unfamiliar terms, too many interactions with frustrating chatbots. We thought that CommandBar solved that problem.

We imagined a future where this search function would become a new universal building block for interfaces. In our Y Combinator application, we described the product as “the natural language interface for any software product.” We were creating our own category (the best name we could come up with was “product search”).

Then we got stuck in the trickiest place for a startup. We found moderate momentum (great customer wins, good revenue growth, strong Series A funding) followed by the realization that it wasn’t scalable (no coherence among customers and unpredictable sales cycles).

The company’s growth only took off when we stopped trying to create a brand-new category and started competing within one—well, two, actually. Instead of focusing on a single way to help users, we now offer a menu of ways to influence users—a “user assistance platform”—that competes in two categories: chatbots and “digital adoption” software (aka pop-ups-as-a-service, or software that allows you to put pop-ups on your website). Our revenue growth rate has doubled, and growth feels much more effortless. We’ve been able to take our original vision—make software easy to use—and connect it to people who care. By labeling ourselves as competitors in those categories, we’re bat-signaling to the very people who feel the pain we strive to solve. We are winning by being less different.

Sponsored by: Every

Tools for a new generation of builders

When you write a lot about AI like we do, it’s hard not to see opportunities. We build tools for our team to become faster and better. When they work well, we bring them to our readers, too. We have a hunch: If you like reading Every, you’ll like what we’ve made.

• Automate repeat writing with Spiral.
• Organize your computer with Sparkle.
• Write something new—and great—with Lex.
• Deliver yourself from email with Cora.

Want to sponsor Every? Click here.

It doesn’t sound as exciting as a futuristic vision of powering the next interface revolution, but it’s a way better business. We show up in market maps (those diagrams describing “10 vendors doing X” that venture capitalists post on LinkedIn), buyers find us more easily, and our demos take five minutes—not twenty-five.

I’m passionate about encouraging other software founders to take the category disruption—rather than creation—path because I think most founders don’t understand they can start that way. The conventional wisdom goes:

1. Great businesses are built on great ideas.
2. An idea has to be sufficiently novel to be great.
3. I can’t succeed by building a startup that looks too similar to other companies.

Debunking this myth has never been more important, because the emergence of generative AI as a foundation technology offers up a more straightforward playbook for category disruption than ever existed before. Here’s how we did it.

Battling incumbents is tough but worthwhile

There are two ways to build a startup:

You can compete in an existing category. Pick a market you think is underserved, build a competitive product, and start competing with existing players. You might not compete with everyone out of the gate, and you may serve only a specific niche within the category, but from the get-go you can say, “We compete in the X space.”

Or else you create your own category. Build something that can’t be shoehorned into an existing market. Maybe it’s a solution to a problem people didn’t know they had or that has just emerged. Or maybe your technology is so new that it can solve a problem people previously thought was impossible to solve. Either way, if you’re in this bucket, you don’t really have competitors. 

You may think that your idea is so unique that it can’t possibly fit into an existing category. But there’s a low-ego reason to be skeptical about being a disruptor within a category, too: You’re competing with incumbents who have more distribution, more money, and more employees than you. Whatever improvement you make, those incumbents can copy. This is especially true in software, where it’s much easier for an incumbent to ship an improvement that wipes out whatever unique feature your startup had. Facebook did this when it saw Snapchat Stories, thought, “Huh, this is cool,” and added Stories to Instagram.

The tarpit of irrelevance

Let’s say I’m building sales software, and I have a point of view about how to help salespeople close more deals by being more organized. Based on the earlier framework I laid out, I have two options: I can compete in the customer relationship manager (CRM) category and build a product that specializes in organization, probably with a clean user interface and some information patterns not found in other CRMs. Or I can create a category by building a sales platform that lives alongside the CRM. 

In order to pull off the first option, you need a product—a differentiated replacement for general-purpose CRMs like Salesforce—and marketing to stand out from the noisier players in the space.

The second option seems a lot less daunting. But it has some critical downsides, particularly in the go-to-market domain:

• Market education: No one wakes up and wants to buy a sales platform. Existing categories possess a sort of buying gravity—everyone knows they need a CRM, every company that reaches a sizable sales team will consider a CRM, people will do market research with the term CRM. A lot of motivated people will find a new CRM company that won’t find a sales platform.
• Buyer education: Even when you get someone interested, you’ll spend a lot of time explaining how you create value.
• Existing budget: The buyer creates a budget for things they expect to pay for. If you need a CRM, are you going to pay for both a CRM and a sales platform? Maybe, if the sales platform can create enough value, but your bar is higher.

So you won’t get squashed by incumbents, but you risk irrelevance.

These problems are exactly those we faced at Command AI. We tried to educate the market in so many ways, but it always felt like we were limited to customers who were explicitly looking for new user interface tools like our search product.

To combat these problems, we instead asked ourselves: “What have we learned about the existing tools that try to solve the same high-level problem we’re solving?” Our conclusion was that those tools were, well, annoying. Take customer service chatbots. They always respond with text, often long passages. That’s pretty aggravating for users to read. We decided to build one that could browse alongside users with its own cursor to demonstrate how to do things, like a tour guide. It was the same insight with a different product, and suddenly we had competitors—but also a lot more people who cared. We didn’t need to educate the market or our buyers, or pound the table for budget. You can see this in how we market our chatbot against old-school chatbot solutions.

Playbook for building an AI-powered competitor

While our product does leverage AI (it’s right there in the name!), we could’ve done so much earlier. Here’s the recipe I would follow on day one as a founder about how to disrupt an existing category.

Step 1: Find the right category

If you stop creating a category and start disrupting one, you’re making a trade-off: You’re presumably accepting competition in exchange for guaranteed demand. But you should look for these warning signs before making the switch:

• The demand isn’t there. If you enter a declining category or cater to yesterday’s trend, you’re getting the worst of both worlds: no demand plus built-in competitors. That’s why the most important characteristic of a good category is staying power. Other companies should have succeeded in the space—and more competition is a good sign! The best sign there’s a billion-dollar company to be built in a category is the existence of a billion-dollar company already in the category.
• Some categories are too “solved.” They’re too hard to enter, too crowded, and the customers are too satisfied. To find a good category, there should be some dissatisfaction with the incumbents. You can usually detect this by talking to people in the space who use similar tools. 

Pick something big (Y Combinator suggests markets that consist of $1 billion-plus of revenue) and has existing players about which people feel kind of “meh.”

You can also ask what kinds of companies are especially susceptible to disruption from artificial intelligence—for example, those with complex user interfaces and are infrequently used (such as human resources software).

Step 2: Build a viable product

You've picked the category and created the dartboard with your competitors' logos. Now you need to build a viable product to replace theirs.

Imagine you’re at a restaurant and you’ve ordered a Caesar salad. The waiter walks over to your table and presents you with the chef’s latest masterpiece: a popsicle made of puréed Caesar salad that, if you dare eat it, tastes just like a Caesar salad—anchovies and all. 

It's easy to be too disruptive and turn your startup into an art project, but most customers want to buy something they understand. Here’s how:

• Build the stuff that checks boxes—even if they’re not particularly cutting-edge. It’s not clear until you speak to customers which features they actually use and which are over-represented in product marketing. So build something that seems viable, market it, do some demos, let your prospective customers point out what you’re missing, build those things (ideally turning those prospects into customers), and repeat. You can’t build a competitive product without the feedback loop of sales cycles.
• Where possible, eliminate functionality that some of the market doesn’t care about. Some of our competitors also make end-to-end customer support software. However, we chose to compete only on the user-facing part—the chatbot itself—instead of the software that lets humans talk to users, too. So we’re competing in a more narrowly defined slice of the category that requires us to build less to be a substitute for others. Another example of something we won’t build is product analytics. We call these features we don’t plan to build “non-features.” 
• As an AI-first company from the start, take shortcuts that your lagging competitors cannot. The most obvious is to lean heavily on prompt interfaces instead of traditional button-and-menu-laden interfaces.

Generative AI facilitates this second step in multiple ways:

• GitHub’s Copilot makes it easier to develop software and make viable products. A viable product is easier to code now that GitHub’s Copilot and other tools can do actual software development. Just like the internet democratized access to knowledge, allowing day traders to have access to similar amounts of information as professionals, AI is universalizing replacement-level engineering ability. As a startup competing with an incumbent, expect to take less time to get up and running with a working prototype.
• AI has made user interfaces less important. Before large language models, every user interaction had to take place with a series of mouse movements and keystrokes with various interface elements, like menus and buttons. Now, a lot of interaction (whether manipulating data or taking action) can take place through natural language interfaces. 
• Prompt engineering is fast. It’s easy to make changes to the prompts that power generative AI features, so startups can move even faster relative to incumbents.

Step 3: Position and market yourself in the category you disrupt

Creating a category means your product is a novelty people pay attention to. Just being another analytics product won’t turn heads.

To disrupt an existing category, you need to tell your customers:

• That you’re part of the category
• What you view as the fundamental problem of the category is
• How you’re solving it differently

Let’s briefly dive into each.

Bat-signaling to your market

It may sound obvious, but to compete in an existing category, you have to say you are part of it. All the time. At every opportunity. Include the category name at the top of your marketing site. Come up with a tagline that references the category: “The CRM that fills itself out.” Ask to be included on every venture capitalist’s market map for the category.

Bonus points if you can leverage your well-known competitors. Our pages that list our competitors are among the most-visited on our site. Saying who you compete with is often the easiest way to explain what you do.

Demonizing the problem

The fundamental problem we were trying to solve was that users dislike digital adoption experiences. For product teams, everyone had accepted that users hate pop-ups, but too bad—we were going to show them anyway. Nobody in digital adoption was willing to question whether users actually liked those experiences.

This is the first step in positioning yourself to disrupt a category: State the problem nobody else is talking about. As a startup founder with fresh eyes, you’re the only one in a position to do so.

Glamorizing the solution

You’ve probably seen a company say something like, “We want to revolutionize how software is built,” only to build a product that looks like everything else. If you’re articulating a novel problem, you should have a credible solution for that problem. This is why AI is so powerful for category disruption: It lets you build differentiated solutions.

Take the AI chat solution we launched. Before LLMs, traditional rule-based chatbots required you to build complex diagrams, even to direct users to an FAQ. Now, LLMs let you point directly to documents and get decent outputs to any related questions. This is a qualitative shift in utility for the user—and your marketing should focus on it.

How do you do that? “We are better because we use AI” may have worked in April 2023, but it probably won’t today. That said, AI ingredient marketing—breaking down the component parts and why they matter—can still work if you credibly advertise the unique facets of AI that will improve a customer’s life or work.

A lot of buyers are open to the idea that AI is cool and makes products better, so if your differentiated marketing message includes claims about how AI makes your thing better, they are probably more likely to land than other claims (all else being equal).

Step 4: Start with disruption, then shoot for the moon

Just because you start off with a disruption approach doesn’t mean you can’t create a new category eventually. Take Stripe. The company started off as a better approach to internet payments, competing against incumbents like PayPal. After dominating that space for many years, it launched a number of brand new features, like Stripe Atlas, which lets customers easily incorporate their businesses. But the foundation was set with a (wildly) successful disruption approach.

The path to greatness often lies through—not around—existing categories where people are already spending money.

To read more essays like this, subscribe to Every, and follow us on X at @every and on LinkedIn.

We also build AI tools for readers like you. Automate repeat writing with Spiral. Organize files automatically with Sparkle. Write something great with Lex.