When running a startup, being able to adapt and change direction is often the key to success. But what happens when you stumble upon product-market fit almost by accident, reach $5,000 in monthly recurring revenue within just 6 months of launching, only to then make the bold decision to pivot to something completely different?
This is the story of Gonzalo Espinoza Graham and Wesley Yue, two engineers who went from running a grocery delivery business to launching an AI-powered lead generation tool called UseDouble.com, before ultimately pivoting to an AI coding assistant named Double.bot.
Along the way, they learned valuable lessons about finding product-market fit, pricing strategy, listening to users, and the importance of solving problems you deeply understand.
Gonzalo Espinoza Graham
Building UseDouble.com
Gonzalo and Wesley started their entrepreneurial journey by running a grocery delivery service in Toronto. One of their biggest challenges was having to manually place inventory resupply orders with different vendors almost every night. With their engineering backgrounds, they knew there had to be a way to automate this tedious process that required a lot of decision-making.
When OpenAI released the GPT-3 davinci-003 model in late 2022, Gonzalo and Wesley saw an opportunity to build a tool to streamline their ordering using AI. Their initial MVP was a simple spreadsheet that could take prompts, feed the information to GPT-3, and organize the output into cells. This humble idea grew into the product usedouble.com.
The product quickly gained attention on Twitter and Hacker News for being a novel application of GPT-3, especially as a tool for lead generation and sales intelligence. Their waitlist grew to over 1,000 people thanks to posts on platforms like Reddit and YouTube. A top 5 finish on Product Hunt also provided a major boost, driving significant traffic without any ad spending.
However, while they had strong initial interest and some loyal users, the product was still very horizontal and not optimized for any specific use case. This led them to the next logical question: “Who would be most willing to pay for this?”
Gaining traction and early adopters
When usedouble.com first launched, they got a big boost from a viral tweet and a timely comment on the popular tech forum Hacker News. At the time, any demo of the GPT-3 AI model was generating a lot of excitement on Twitter, and the founders were able to capitalize on that by highlighting their spreadsheet-based tool.
The initial product they offered was pretty straightforward - it allowed users to feed information from their spreadsheets into GPT-3 and have the output organized right in their spreadsheet cells. As they talked to these early users, the founders realized there was also a need for a web scraping tool, since many users weren't familiar with the concept of AI language models "hallucinating" information and the lack of context from just scraping the web.
The combination of the Twitter and Hacker News buzz, along with the unique features of usedouble.com, helped them quickly build up a waitlist of over 1,000 users. The company's growth was further fueled by strategic posts on Reddit, listings on AI tool aggregator sites, and YouTube videos.
They also had a very successful launch on the popular Product Hunt platform, where they secured a top 5 spot. This brought them even more exposure, as Product Hunt amplified their presence through their social media and email newsletters.
So in summary, usedouble.com was able to gain early traction and users through a mix of social media, community engagement, and strategic marketing - capitalizing on the hype around new AI technologies like GPT-3.
Niching down and monetization
As more people started using usedouble.com, their costs for the GPT-3 AI model also grew rapidly - from just $7 per month in December 2022 to over $500 per month by February 2023. It was clear they needed to start monetizing the product to cover these increasing expenses.
After talking to a lot of customers, the founders Gonzalo and Wesley realized that one particular group had the biggest need and was willing to pay for their tool - B2B sales teams. So they set out to build the best AI tool specifically for salespeople.
They added features to help sales teams find and verify email addresses at scale, enrich their lead information with relevant data from across the web through simplified workflows, and integrate the tool with their existing CRM systems. One especially popular feature was the "Online AI Instruction", which could intelligently search the web to try to answer specific questions about a prospect or company by pulling information from sources like LinkedIn, Crunchbase, and Google Maps.
Eager to tap into the strong demand from sales teams to automate their prospecting, usedouble.com started charging up to $500 per month - a significant jump from their original plans to offer the tool for $20/month. Surprisingly, this higher price point actually attracted more serious and engaged customers compared to the lower-priced plans. They had found an underserved niche and built a product that truly solved a real problem for these sales teams. This business model paid off, as they grew to $5,000 in monthly recurring revenue within just 6 months.
Challenges and Insights
Despite the initial success, usedouble.com faced some challenges. There was constant demand for more data sources, but the low differentiation and pricing power of the data limited their potential for growth. They also found that users would often import a list, enrich it, and then not come back for months. This suggested the platform wasn't solving an urgent, recurring problem for most of its user base.
The founders also discovered that most users required simple enrichment, such as phone numbers, emails, and social media handles, which could be better served by traditional methods rather than LLMs. The demand for long-tail enrichments was less common than initially anticipated.
Experimenting with outbound campaigns
Gonzalo and Wesley, started wondering how they could go even deeper to solve more of their users' core workflow. They wanted to keep their customers engaged and deliver even more value.
After doing some user research, they honed in on a key insight - enriching lead lists was only the first step. The natural next phase was helping their customers launch outbound marketing campaigns to those prospects.
Their hypothesis was that by using AI to personalize emails based on the rich company and prospect-level insights they had, they could generate much higher response rates compared to generic email blasts. They decided to test this theory by running full-service outbound campaigns for a select group of enterprise clients.
For 2 months, they used their own product to create highly customized emails for over 10,000 prospects per week. However, the results were disappointing - their conversion rates were no better than the basic templated emails the customers were already using. The salespeople actually preferred to pair the lead enrichment data with proven templates rather than untested hyper-personalized messages.
This was a pivotal moment for Gonzalo and Wesley. They realized that even with advanced AI, it's extremely difficult to accurately infer whether someone is a qualified buyer based solely on public information. Trying to hyper-personalize outreach solely based on sparse signals was not the game-changer they had hoped for. They realized the real breakthrough would need to be something that more directly impacted sales workflows.
The pivot and the birth of double.bot
Around the same time that Gonzalo and Wesley were exploring how to better support their sales team customers, they were also experiencing their own frustrations with existing AI-powered coding tools.
As avid users of GitHub Copilot, they were surprised to find that many of the most highly upvoted feature requests and bug reports had been ignored for over 2 years in some cases. They saw a huge opportunity here to build something better - a more reliable and responsive AI programming assistant.
Leveraging their own deep knowledge of the developer experience and the latest AI models, Gonzalo and Wesley got to work building a new tool called double.bot.
Double.bot
The initial response to Double.bot was overwhelming. They gained 10,000 users within the first 2 weeks of launching the product - the same amount of users that usedouble.com had attracted over the course of 6 months. This further validated their decision to focus their efforts on a problem space and user persona that they intimately understood as engineers themselves.
Key takeaways:
- Don't be afraid to charge more: Gonzalo and Wesley found that moving upmarket from $20/month to $500+/month attracted higher quality, more engaged customers who provided better feedback.
- Listen to users, but know when to ignore them: User input is important, but it's not always productive to completely shift your value proposition or strategy based on feature requests. The founders spent months chasing a personalization hypothesis that didn't pan out, even though some users were pushing for it.
- Solve problems you deeply understand: The rapid adoption of double.bot showed that Gonzalo and Wesley were much more effective at building for the use cases and needs of developers (like themselves) compared to sales teams.
- Incremental growth adds up: While they benefited from some viral posts, the majority of their growth came from less glamorous channels like SEO and content marketing. Slow and steady efforts compounded into a meaningful acquisition engine.
- You can build a real business by accident: Despite being initially focused on grocery delivery, a simple prototype spreadsheet ended up attracting the most engaged users and growing into a revenue-generating tool. Spotting those signals and pursuing them was key.
Conclusion
In the end, what started as a homegrown automation tool for their sales team turned into a promising SaaS business in the form of usedouble.com. But Gonzalo and Wesley came to realize that their greatest advantage actually lied in building tools for their own kind - engineers.
Pivoting from lead generation to an AI coding assistant like Double.bot was a bold move, considering the traction that usedouble.com had already gained. However, it played much more directly to the founders' own strengths and experiences as engineers.
Only time will tell if Double.bot is able to reach the same revenue milestones as usedouble.com did. But by applying the hard-fought lessons they learned from their first accidental success, Gonzalo and Wesley have significantly increased their odds of building something even bigger this time around.
