Death by AI - Jumping on and off the AI Hype Train 🤖 🚂
If you aren't using AI you will die
This and similar messages pop up on my Linkedin every day.
You cannot ignore the boom happening in Generative AI, obviously all your follows are posting and reposting 'top AI tools', 'AI news', 'dont miss out on this and that AI'.
My years in AI
Being in the middle of AI and working with Gen AI for the last 4 years (Colossyan, Defudger), I am really fatigued, as you can imagine. But, as it happens, I left my Gen AI company (that just raised 5 million dollars). Not the happiest story, but that's for another post to discuss.
Nevertheless, I wanted to stay close to AI, however tired I already was of it...and jumped on the ChatGPT hype train, coming up with the idea for Upsum.io. Having worked with generative AI for four years, I was there to observe the advancements in LLMs (large language models). While many founders were focusing on using AI for content generation in marketing and social media, I saw an opportunity to utilize the models in a different way.
Here we go again
The inspiration came when I read through an edition of the Microsaas newsletter that highlighted the untapped potential of leveraging GPT models for content analysis. The concept of creating a summarization tool that could help individuals efficiently read through lengthy articles, YouTube videos, and podcasts, resonated with me. I envisioned a tool that could be invaluable in professional settings, enabling high-paying employees (consultants, lawyers etc) to quickly grasp the contents of contracts, reports, and other learning materials, thus boosting productivity.
Excited by the idea, I set out to build Upsum.io, with as little effort as possible. Knowing the hassle of working with databases, AWS, hosting etcetc, I did a week research to see what are the nocode alternatives, that can make my work faster and easier.
I was able to develop the initial functional minimum viable product (MVP) within just a week. Using Bubble and integrating the ChatGPT API for the AI text summarization capabilities, I was able to bring the product to life quickly and efficiently, happy with myself and the amount of effort and actual work it required to build Upsum 1.0 — with the MVP ready, we started focusing on marketing with my cofounder (non-tech, he was with me for the last 4 years).
The first 10 users
With all the noise and short attention span (thanks Tiktok) of people, it is hard to reach people. On the organic side, we spammed discussions on platforms like Reddit, where we answered threads related to content analysis and summarization. I also posted about Upsum.io on forums and websites that catered to the AI tools and technology community. By reaching out to these communities, we got a couple of visits on our website and collected emails for free summaries.
Personally, I find Google Ads pretty useful where there is existing demand, meaning ppl search for your keywords, so that you can harvest that. This allowed us to reach a wider audience and bring in a steady amount of new users every day on a USD 10 daily budget. By carefully optimizing the campaigns and targeting keywords relevant to content analysis and summarization, we got 100-150 new unique free users every week.
Going PRO
Monetizing Upsum.io followed a product-led growth (PLG) strategy. Initially, we offered a free trial of the tool, allowing users to experience its value without any financial commitment. This helped us gather feedback and insights from users, enabling us to improve the product and address any issues or limitations.
After a month, I introduced a Pro version of the tool, which offered additional capacity and advanced features to users. By listening to the needs and feedback of our free users, I was able to create a paid version that catered to their requirements (still in progress). This approach proved successful, as within three weeks of launching the Pro version, I converted 10 free users to paid subscribers, right now, after roughly 2 months, we have a base of users and we can collect feedback on what new features to add. We will go though a redesign of the website and app with the help of a designer and will focus more on appearing on those tedious 'Best AI tools' posts on Linkedin and Twitter
— Attaching some screens from our first Miro board —
ChatGPT, please insert conclusion:
In conclusion, my journey with Upsum.io has been a whirlwind of innovation, growth, and customer-centricity. From the initial idea to building the product using no-code tools, marketing it through various channels, and implementing a PLG strategy for monetization, each step has been a learning experience. I am excited to continue evolving Upsum.io and exploring new ways to help users read through long documents efficiently with the power of AI.
Personally, I'm totally exhausted by the current AI hype train. I've stopped using twitter because of that, and I'm more reluctant to read any news, because everything needs to have "AI" in it. Plus the way how most of the trending messages are structured (as you mentioned in your post) is almost depressing - like all my of my "For you" section on Twitter is the list of what "you're doing wrong".
So my answer is No - I'm not. But congrats on your first paying customers!
p.s. I do use AI - I find it sometimes useful, sometimes frustratingly bad, but all in all, a tool, no hype
Agreed, completely :D AI is a great tool for some uses, and horrible for others! But, it will get better and better...
Many girls who go for permanent makeup for the first time are afraid that the pigment will not take root. Such fear often arises due to the fact that the masters do not tell what happens to the lips after the introduction of paint.
Hope you enjoyed trying the tool ;)
I like how you offer a free trial on the landing page and (I assume) use it to capture users email to nurture them. For advanced usage, they can upgrade to Pro - smart!
I assume 99% of your users will stay on the free plan - do you incur cost to support them? (i.e. api calls to OpenAI)
Right! We are really good at visitors who try the tool (about 40%) and we are around 2% conversion from visitor to paying user.
We are currently supporting everyone, the main use of a free user is to learn from them and make the product better.
great!
If you don't mind - how do you think about future marketing channels.
Unit economics for Ads might be a challenge. I can see you're paying like $1 per click. E.g. If you put $1,000 into google ads, you get 100 users. 2 of them will pay (say 10 dollar per month and they churn after 6 months (wild assumption)). Your ROAS is like ~0.1 (2x60/1000) then.
But I guess there are possibly low hanging fruits for programmatic SEO (text summariser for industry1, text summariser for industry2; text summariser for use case1, text summariser for use case2) (koala.sh has some nice api now) and maybe affiliate. I am also currently considering building a product for the German market only - the SEO/Ads competition is much lower, I feel.
As you say, correctly, Ads is a good way to get early traction and the first users, we need to work more on organic channels and SEO, possibly building a community.
Also, we are thinking on putting some focus on the long-tail and work with Spanish, German, French text.
I wonder do you have any plan to keep it alive against ChatGPT plugins? I believe building ChatGPT-based products will be hard in following days since it won't be possible to compete with plugins, unless you have budget to compete . There is already two plugins on ChatGPT to summarize PDFs. It's for paid users, but it will enhance.
Sure, it is a threat. We will try to find niches, and target customers who will need a private and secure solution, so they wouldn't wanna upload private documents to ChatGPT. Additionally, ChatGPT is 20$ a month, so we will price accordingly, someone who only needs ChatGPT for a certain task, will be happy to pay 10/ month for that, I imagine.
What do you think?
It makes sense. But pricing is hard in that case. Tokes costs of some users basically would exceed 10$/month. Probably there have to be some daily/montly limits and users may not like it. I believe it's a challenge to keep going with apps that only depend on being a ChatGPT wrapper. I believe such apps should provide more than that to succeed.