Copy.ai – the AI platform for go-to-market teams – launched in two weeks using just seven software tools. Within three months, they got over 40,000 users. Now they’re celebrating an unreal milestone — 10,000,000 users (😱).

I sat down with Founder & CEO Paul Yacoubian to ask the question, how did that happen?!

Paul was extremely candid in our conversation and I’d like to let you hear from him directly. Keep reading for the highlights and don’t miss the video clips mixed in throughout.

Share

“If you build an innovation but can’t distribute it, that innovation is worthless to humanity,” Paul told me.

He had the idea that there must be some way to “copy and paste” businesses processes from one company to another, and he lasered in on go-to-market processes specifically. But there wasn’t a scalable or technology-based solution for doing that.

The way that business processes got imitated historically was that you’d hire someone who did X at their past company to come do it again at your company. Or you’d find a consultant who would bring in best practices from similar companies and help you implement them. Only the top 1% of companies could afford to do this, Paul hypothesized.

He wasn’t able to do anything with this idea until all of a sudden OpenAI launched GPT-3 in beta in July 2020. “It was the most mind-blowing thing,” he recalled. “It would consistently create something valuable for you. If a company could talk back to you, could you sell the words it’s producing?”

Share

Paul got excited about the potential he saw with GPT-3. The question became how do we validate that use case as fast as possible?

His approach: build an MVP in a matter of weeks, launch it on Twitter and get feedback to keep iterating. The fifth launch was Copy.ai.

“There was a flywheel where when we’d launch, we’d get more followers and more people would be interested in what we’re doing,” Paul said. “And so the next time we’d launch, we would have even more of a kickstart to that project.”

He focused on Twitter because that’s where the early adopters were and they were there in real-time. Within three months of launching the fifth product – Copy.ai – they got over 40,000 users. This traction was noticed by VCs, too, and within three weeks they had a seed round.

While this rapid iteration might seem unattainable, Paul emphasized that Copy.ai did it using only seven software tools. “Before we started building, we didn’t know it was that easy. But you could get a great company off the ground with a minimal stack in 48 hours. It only costs $100.”

Specifically, the team used Webflow as both the homepage and as their app (not recommended, Paul said). They had Magic (authentication), Firebase (backend), Stripe (payments), Notion (project management), Slack (communication and alerts) and their own product Copy.ai (copywriting).

The Copy.ai team started monetizing the fourth MVP, a marketing tagline generator called taglines.ai.

“We just wanted to see, would anyone use this? And would they come back and use this again? And if we charged them $3 a month, would anyone pay us?”

From there, Paul tested raising the price to see if someone would pay $5 a month. They did.

Then he raised it to $10 a month. Customers kept paying. But this time, they started taking the product more seriously and adopting it for other marketing and sales use cases.

Armed with this insight, Paul went after a broader platform opportunity around these tools. He bought the domain Copy.ai and was off to the races (Paul even remembers that it cost $6,000 at the time, but he could rent-to-own it for $167 per month.)

If people would pay $10 per month for a single use tool, what would they pay for a platform tool?

Paul started to consider the alternatives. On the high end, a customer might pay $2,000 per month for an ad agency – although he recognized the tool couldn’t replicate that experience right away. He figured that prosumers or smaller customers would prefer a DIY experience as long as the cost was “a lot less” than the ad agency.

Share

“We kind of threw out a $49 a month number with a discount for an annual plan,” Paul recalled. “And then it just worked. Straight out of the gate.”

He did ask users about different price points. In doing so, Paul learned that a $49 price wasn’t a consumer-type price and was oriented toward a recurring user. The economics worked at this price point, too. Interestingly, he has continued to test different price points and found that the experiments have always been worse than the control.

Initially, Copy.ai had a free trial model where customers could use the product unlimited for seven days. As costs for compute have gone down, Copy.ai has been able to introduce a freemium offering with a usage limit. Today’s free plan comes with 2,000 words and 200 bonus credits.

It’s an understatement to say that Copy.ai grew quickly out of the gates. The product hit 2,700 users in the first three days and 40,000 in the first three months.

The company’s early users came from Twitter, then other social media platforms. “You had this moment in time where people hadn’t used generative AI yet,” Paul admitted. “That’s the opportunity, and then you arbitrage the CAC economics to get people to come in and try the product.”

Copy.ai later invested big in SEO optimization (the product is for content generation, after all). Paul looked at the SEO playbooks that had worked well for other SaaS companies. One of these was free lead magnets such as HubSpot’s blog title generator or Shopify’s business name generator tool (I wrote about these as part of a product-led marketing strategy).

Copy.ai built out a lot of these – by my count there are at least 30 on the company’s website. (“Apparently people need help generating Instagram captions,” Paul quipped.)

These pages attracted a lot of traffic for Copy.ai. And there were compounding benefits as Copy.ai earned domain authority and ranked for more SEO terms. This gave Copy.ai an advantage as they started going after terms more related to enterprise decision makers. “All this stuff is a flywheel and none of it is a funnel,” he told me.

Paul contrasted his approach to what he saw from competitors. While a number of competitors went into stealth mode out of the gate, Paul was building Copy.ai in public and pushing hard on distribution. He sees this as core to Copy.ai’s success to date.

“In new tech adoption cycles the way they usually play out is that it takes two to three years for incumbent companies to actually add that and bring it into their existing products. So you really have a two to three year head start. Then you have to figure out how to build distribution for yourself in that time horizon. It’s really all hands on deck.”

- Paul Yacoubian, Copy.ai

Scaling further meant pushing on the channels that were already working.

In the early days the “marketing team” was really just Paul on Twitter. To get to 10 million users, Copy.ai hired individuals to specialize in social media, SEO and lifecycle email. (That’s an extremely lean marketing team given the company’s trajectory.)

“You don’t need many people,” Paul advised. “It really helps to be able to focus a person full time on one channel. I would recommend that for any budding founders out there.”

Copy.ai had a prosumer model, which lent itself well to social media. Paul has noticed that most enterprise SaaS companies tend to imitate what they see and there are very few personalities in B2B. That creates an opportunity for anyone willing to put themselves out there. “Most of the time people are eager to learn and they like to be entertained… If you’re a founder, people want to be inspired by you.”

Share

This post might be cut off by your email provider. Keep reading the full version here.

Much of Copy.ai’s distribution was enabled by AI. Remember those lead magnet tools for SEO? Copy.ai uses AI to:

Generate the tools
Generate the descriptions of the tools
Generate the landing pages to promote the tools
Generate the examples going into the tools
Generate examples of inputs and outputs

All that Paul’s team has to do is review the drafts and add light edits.

The way Copy.ai uses AI now has changed from when they got started. For example, the team still uses AI to write articles, but starts with a conversation that already happened like an internal call or a call with an expert consultant. “We get ready to go blog posts without doing any work… the whole thing that makes the content flywheel spin is an original conversation with net new information.”

Another AI use case is in Copy.ai’s sales process. When someone signs up for the product or requests a demo, Copy.ai runs a real-time AI workflow. This workflow researches the relevant people, their roles and the company itself to create a multi-dimensional picture of the target customer. Copy.ai’s go-to-market teams then leverage that picture to personalize every downstream interaction including:

The inbound sales email sequence
Recommendations for use cases that might be relevant
An account plan for that company itself

“Basically everything downstream in that sales process is getting massively expedited with really great data driving it.”

Paul contrasted this approach with the historical mental model where a company is targeting a persona or an industry. “You're getting really granular about the actual individual,” he said. “And with that level of granularity, you can create amazing customer experiences for them.”

I had to ask Paul about what product-led growth (PLG) means for Copy.ai and if any PLG initiatives helped inflect growth.

The first initiative that came to mind was the sign-in with Google in one tap pop-up.

“How easy can you make it to sign-up? I think the easiest would be if you don’t need to sign up at all. But the costs are really too prohibitive for that.”

The second was introducing freemium. Paul had realized that sometimes people would find the product, but they weren’t in the buying motion yet and were still in the discovery phase. With a seven-day free trial, the user would lose access to Copy.ai before they need to truly evaluate it.

“There’s a lot of research you can do ahead of time before you have a sales call. I think designing for that when you go into the enterprise is really key.”

Paul emphasized that when a SaaS company sells into the enterprise they encounter a buying group with five different people on average. These people are at different levels in a company from end users all the way up to the economic buyer. This means there are five different potential entry points into a company. “I see companies make a mistake where they just say ‘OK, we need to pick one and just go after that person.’ But you really have five shots on goal.”

These end users can be powerful and overlooked advocates in an enterprise deal. “Most of the time, they will literally just show you how to win the deal. They’ll say, ‘These are the four people you need to know and this is what they care about and this is the order you need to go in.’”

Share

We wrapped up the conversation reflecting on Paul’s advice for anyone thinking about building an AI startup.

“You’re four years too late,” he deadpanned.

Paul was referring to the pivotal Attention is All You Need paper, which has been cited by 110,000+ people according to Google Scholar. “If you look at where the dollars ended up, it has to be north of 90% of the AI capital raised had their name on the paper,” Paul said.

On a more serious note, Paul recommends building a bootstrapped app that’s vertically focused. And to try to own as much of the value stream as possible for that application.

“Treat is as a service. If someone wants to buy the service, it’s not just the software, but it’s operating it. You could rinse and repeat that model across every industry and probably every department in that industry as well.”