How ChatGPT saved my startup 🥲

How it started

On Christmas Eve 2020, I decided to quit my well-paid job at Atlassian to do my own startup again. The idea was to build a new kind of app (called Questmate) that would revolutionize how individuals and teams of any size would get things done.

The general concept of the app was to allow users to create so-called Quests, which would not only entail the steps needed to be done for a particular task but also allow you to inline guidance and actually enable users to do things right from said Quests— making it incredibly easy for everyone going through the Quest to get it done fast, and with confidence & actual joy. 😇

Quickly after building the initial prototype, I went on to raise a US$1M+ pre-seed round using nothing but a Quest. From there, I assembled a small team around me, and together we progressed further in filling all the missing pieces in core functionality and started to onboard our first customers, big and small. 🚀

And guess what? We did it. All of what we planned is now possible using our app. Everything from either sharing Quests publicly or directly assigning them, scheduling them, adding reminders and optional review steps on completion, to automatically issuing rewards (cash payout, gift cards, etc.) for finishing a Quest... all available on Web, iOS & Android, completely passwordless and with 100% feature parity.

We also developed a way (including the service powering it) that allowed us and our customers to build all the “functional” items as part of a Quest — allowing users to do everything from controlling a Philips Hue light, disarming an alarm system, unlocking and opening a door, checking in participants in another app, to starting a car. All that directly from the Quest and without ever leaving the app. Every aspect of “getting things done” is covered by the Quest, from initial notification to things being done and the person doing it being rewarded. 🤯

Today, Questmate allows users to create everything from simple checklists to fully-fledged forms, workflows & interfaces for their daily routines and tasks.

Here’s an idea of what such a Quest can look like, as per example of a Quest to cover a class at a local Yoga Studio.

Even better, we acquired a good bunch of actual, paying customers using all this functionality daily. For everything from checking in visitors, renting out and giving control to Teslas, collecting feedback, managing deliveries, selling and managing tickets, all the way to running their routines like shop opening/closing routines, team standups, or deployment checklists. 😎

Time to pop some champagne, right? 🍾 Well…

The Challenge 😳😬😏

Unfortunately, at the beginning of 2023, we were still acquiring and onboarding most of our customers manually, following the well-known “do the things that don’t scale” principle.

So while we kept on shipping improvements and new cool things for our existing customers like crazy (leading to an almost 100% retention rate to this date), the very few new users coming in organically (you know, via Google Search and doing social posts) unfortunately almost all never made it to become an active, paying customer.

Almost awesome, one could say!

ChatGPT to the rescue 🦄

Having been an early user of ChatGPT (the app), I pretty much tried to apply it daily to the problem we had at hand. For creating some of our content, generally assisting in figuring out potential things to try, etc.

And while implementing ChatGPT into our app itself was obviously always on my mind, I always thought this should come after we figured out and solved our user acquisition. But then it suddenly hit me…

Why not solve our user acquisition challenge by allowing users to create Quests using ChatGPT before we even ask them to sign up for our app and actually use them? That way, we would learn what users actually wanted to get done while providing them with value… and all before even collecting their email addresses.

My initial idea was to change or replace our existing landing page or just create a second one from scratch. Fortunately, the idea received quite some hesitation and fair pushback from the team, mainly because of the distraction it would create from serving and keeping our existing customers happy and retained. That’s when it hit me yet again… ⚡️

“Why not build the MVP the same way we usually do when trying things at first? As a Quest itself!” 🤩

So the very same day and moment, just 1 hour before a talk I was invited to give at a local startup event that evening, I started building a simple Quest that would serve this experience (User Prompt -> Quest Template creation) while in parallel harassing Jack (our engineering lead) to quickly register the domain Doogle.app which we would redirect to the Quest and would be featured on the last slide of my presentation that evening. 💦

And fast-forwarding two weeks from that night, this is what happened to our user-growth curve…

Pretty sweet curves, eh? And what if I told you that the same thing — just bigger — happened again when we got our invitation from the fine folks at OpenAI to create and publish our plugin directly on ChatGPT with approximately +100M users to date? 💥

Now, you might be wondering. How many of the users we ever show an autogenerated template to, based on their prompt, would actually hit that “sign me up!” button? You’ll probably don’t believe me, but it started with some jaw-dropping 25% and just recently hit ~40% 🤯 I kid you not.

And we didn’t even start improving the actual Quest creation process much. In fact, it’s still running on a GPT-3.5 turbo… not even GPT-4, which would create even better results. 🙂

How it’s going

“So that’s it? You’re out of the wood?! Flying all high? Finally popping those champagne bottles?”

Well, at least we hit the local arcade lunch together when we made our first inbound sale the day our first “Prompt to Quest” experience went live 🤣

There is still quite a big gap between the users coming in these channels vs. the ones we still acquire and onboard manually, e.g., via our outbound sales pipeline. That’s why we currently focus on improving the conversion rate from these new users to sticky, paying users before creating more hype again and extending our top-of-funnel. Well, apart from this blog post, that is :)

Conclusion

Depending on where you are in the journey, it might be worth considering adopting ChatGPT to solve your top-of-funnel challenges before applying it left, right, and center inside your product. 🍭🍬

Make sure to check out our ChatGPT plugin for inspiration, or — should you not have access to ChatGPT plugins yet —simply try our Prompt to Quest creation via Doogle.app.

You can also hit me up on LinkedIn or Twitter (well, X 😮‍💨), as I’m always happy to share more and help wherever I can.

Until next time ✌️

P.S. If this post gets enough likes, I’ll follow up with a more technical post and even share some of the code that’s powering the Doogle.app and the actual Quest creation. Spoiler: It’s not rocket science ;)