TLDR: we built a little consumer app in a day. It went viral, bringing us leads, but more importantly, it allowed us to become users of our own product, Dawn.
Everyone in SaaS should be trying this.
How we did it👇
A Small Viral Experiment:
A few weeks ago, we realized we had a little problem.
We're building Dawn, analytics for AI products, especially conversational ones. But we aren't a conversational AI product.
This can be a disadvantage. So a few weeks ago, we started looking to fix this.
Inspired by Wordware‘s amazing Twitter roast app, we came up with a plan.
We'd build a small conversational AI app (called ChatGPTPrank), hook it up to Dawn for analytics, and make it go viral. After finalizing the app, we posted it on Twitter, Reddit, + Hackernews - and it went viral almost immediately, getting us tons of inbound leads.
The best part: The whole thing took < two hours.
Building the app (<1 hour)
With AI + serverless, it’s never been easier to build and ship something.
We went from idea to live in just a couple hours.
We used Vercel v0 for the initial code + design, plus Cursor to finish the job. And we bought the domain + deployed on Vercel, running the backend on Upstash (serverless Redis).
Lead Gen
Our app is called ChatGPTPrank - it's like LetMeGoogle that for you, but for AI.
It was critical that the site's virality drove website traffic for Dawn itself. So, we added CTAs in the site that linked to our dawnai.com page - things like "built by Dawn", etc all hyperlinked.
Visits to the homepage jumped more than 1000% and are still elevated, and we got tons of good leads, and all this was with hardly a few hours of work.
Dogfooding
The most important part - using Dawn.
This was the experiment’s main goal - to have a conversational AI product that allowed us to be real users of our own product.
Dawn helped us monitor how our AI product was doing - to find trends in usage - what issues people run into that cause them to churn, what topics that resonate that lead them to stay.
Interesting findings in the ChatGPTPrank data:
— Surprisingly, people use it for serious stuff - like business ads or political debate
— Lots of people use it to roast friends or sound like Donald Trump
– And fully 60% of people who get pranked themselves start making prank links (good viral coefficient)!
The Painful Learnings for Dawn
We realized a few things from being our own users (of Dawn)
— Latency matters: any page, chart or table that felt slow in Dawn felt painful. We always knew this logically, but now being our own users means we feel the pain. Speed things up.
— Great overview pages are essential. We realized that the second you log in to dawn, you should be able to see all your metrics in one place - the pulse of your AI app. It should feel addictive. So we redesigned our app around the overview.
— Issue detection matters: What matters are the issues in your AI app, not just what people use it for. Having a huge number of visitors is awesome - but most importantly - where is your agent messing up, where is it leading to user frustration, or worse, churn?Â
So, we are zeroing in on helping customers find, monitor, and fix issues in their AI product — like Sentry for AI products.
Conclusion
You can just do things.
Not your own user? Ship something that makes you your own user.
Need to boost your lead gen? Try making a little app that goes viral.
Things have never been easier.
Zubin Koticha is the Founder of Dawn, product analytics for AI products.