⚡️ Competing with ChatGPT and Sierra, building a $10M ARR company — Yasser Elsaid, Founder, Chatbase
TL;DR
Chatbase hit $10M ARR without raising — Yasser Elsaid says the company went from launch to $1M ARR in 117 days, stayed bootstrapped, and now runs with a 24-person team competing in customer-service AI against far better-funded players.
The original wedge was dead simple RAG before RAG had a name — In late 2022, Yasser saw GPT-3 plus custom context as obviously valuable, dropped his other projects, left school on pause, and shipped a website chatbot trained on company docs to his 16 Twitter followers.
Virality mattered, but 'doing the normal business stuff' is what made it real — Yasser credits sales calls, demos, user conversations, and relentless follow-up—not just a viral launch tweet or LinkedIn hype—for turning an AI demo into a durable product.
Chatbase’s edge is PLG first, not sales-first enterprise theater — Instead of 'call us' gates, the company built a no-credit-card self-serve flow and now uses that product-led motion to pull in customers from Chuck E. Cheese to the UN, then layers sales help on top.
In customer service, the bottleneck is mostly the harness, not the model — Yasser estimates 95% of the limitation is orchestration, guardrails, and workflow design, and says with the right harness you can reach 80–90% automated resolutions in many industries.
The roadmap is bigger than support: Chatbase wants to be a 'chief customer officer' — Yasser’s next step is an agent that handles support, sales, and onboarding across channels while mining every conversation for product feedback, churn signals, and growth opportunities.
The Breakdown
From Canadian student side-project machine to accidental AI founder
Yasser starts with the classic indie-hacker origin: a computer science student in Canada, 20 to 30 side projects deep, trying to make money online while doing internships. He was originally building essay-writing tools for students, but once he started playing with GPT-3 in late 2022, he realized that adding custom company data to a model—what we now call RAG—was the real opportunity.
The tiny launch tweet that caught the wave
He dropped everything, stopped going to school for a stretch, and built the first version in about a month and a half. The launch went out to just 16 Twitter followers, but the demo—basically ChatGPT with your own docs plus source chunks and guardrails like refusing to answer 'what’s the speed of light?'—felt fresh enough to go viral before he even had pricing.
Why Chatbase didn’t die after the hype cycle
Yasser says the difference between a flashy AI demo and a real business is boring in the best way: sales calls, demos, talking to users, and actually shipping. He failed his last semester during the sprint, maxed out a $25,000 personal credit card paying OpenAI bills, and had customers yelling when payments failed—but that grind is what got Chatbase to $1M ARR in 117 days.
The early product: 'just upload docs and embed a chatbot'
The first product wasn’t some giant enterprise platform; it was a simple external-facing chatbot for business websites, trained on company documentation. That clicked because companies were used to rigid drag-and-drop support bots, and suddenly here was something that felt like it could actually change customer support—then later expand into internal use cases like Slack.
Staying bootstrapped while everyone else raised on vibes
He says he intended to follow the normal startup playbook—pitch deck, cofounder, investors—but traction came too fast, so he just kept building. Now Chatbase handles huge model volume across OpenAI, Anthropic, and Google; Yasser says around half of usage is still OpenAI, but Anthropic and Gemini are growing, and switching models isn’t free because customers spend months tuning behavior and guardrails.
The real product problem: the harness, not the IQ
One of the sharpest points in the conversation: for customer service, Yasser thinks 95% of the limitation is not model intelligence but the 'harness' around it. The models are already smart enough, he argues; the hard part is orchestration, instructions, escalation logic, and making the system reliable enough to reach 80–90% resolution rates without breaking handoffs.
PLG as strategy, not just pricing page aesthetics
Chatbase grew differently from competitors by being aggressively self-serve first: sign up, no credit card, clear pricing, try the product now. Yasser frames that as a forced advantage of bootstrapping—he couldn’t hire 10 salespeople on day one—yet now it lets them attract 10,000 active customers and discover weirdly valuable segments like theme parks, Chuck E. Cheese, F45 Hong Kong, Opel, and even the UN using Chatbase on Facebook Messenger in crisis regions.
The go-to-market machine and the bigger ambition
Yasser’s GTM playbook is blunt: if you’re B2B, do B2B stuff first—demos, outreach, cold email, warm outbound—then layer content on top so every channel works better. Looking ahead, he wants Chatbase to evolve from support bot into a 'chief customer officer': an AI brand ambassador that handles support, sales, and onboarding conversationally, while surfacing the gold mine hidden in customer conversations to tell companies how to improve the business itself.