Staying connected with AI
How thoughtbot validated and built a proof-of-concept mobile app using AI to tackle the challenge of maintaining social connections
![Knect website with text "A delightful way to keep in touch with the people who matter most".](https://images.prismic.io/thoughtbot-website/ZnROq5m069VX17R1_framed-hero.png?auto=format,compress)
“Coming out of it, I now have a specific understanding of what would be valued and how by specific people. Now I know I have a good sense of what to actually do.” — Josh Herzig-Marx, Founder of Knect
![Screenshots of the Knect mobile app with recommendations of people to reconnect with.](https://images.prismic.io/thoughtbot-website/ZnRT15m069VX17Vm_3-screen.png?auto=format,compress)
The Challenge
As a solo founder, how do you determine if an AI opportunity that you identified is technically feasible and worth pursuing?
With over 20 years in product management and tech, including prior endeavors as founder and a successful acquisition to Google, Josh Herzig-Marx is a seasoned entrepreneur.
This time, Josh was on a mission to tackle a challenge many of us could relate to – staying connected. He had been working on a solution (Knect) to address the difficulties we encounter in maintaining connections with people, especially weak social ties that can quickly slip through the cracks.
His vision involved a solution, potentially leveraging AI, that would passively keep an eye on people’s communications, help identify their most important relationships, and support users to stay in touch with them.
Josh came to thoughtbot to validate that this was an opportunity worth pursuing and learn whether it was technically feasible.
![Two women chatting in a professional context.](https://images.prismic.io/thoughtbot-website/ZnFZVJm069VX11dA_microsoft-365-FHhbHW4vFxc-unsplash.jpg?auto=format,compress)
What we did
Finding message-market fit through discovery sprints and rapid prototyping
We kicked off our work with several Discovery Sprints to validate the opportunity and find the right market segment. Our first goal was finding whether the pain point was widespread enough, as well as whether people cared about it to the point of paying for a product.
The thoughtbot team worked closely with Knect founder Josh, in conducting research and discovery work including user interviews, competitor analysis, market segmentation exercises, assumptions mapping, and user testing.
![Screenshot of the Knect mobile prototype showing the inbox of recommendations to connect with.](https://images.prismic.io/thoughtbot-website/ZnRRHpm069VX17UJ_Inboxhighlight.png?auto=format,compress)
In order to develop the solution further and assess message-market fit, we iterated on the product through rapid prototyping, validating and evolving our ideas with user feedback during the interviews with potential customers.
This discovery work informed the technical strategy that we were about to tackle in the next phase of the project. Before we invested substantial time and money into building the actual product, we made sure to establish the foundations of our product strategy by learning what would provide real value to users.
![Screenshot of the Knect mobile prototype showing the a recommendations to connect with.](https://images.prismic.io/thoughtbot-website/Znl--ZbWFbowezqy_Profilehighlight.png?auto=format,compress)
De-risking through technical feasibility spikes
With a clear idea of what was most valuable to our target audience, our next goal was outlining an architecture to clarify what could be built and get more specific on how difficult it would be.
In a series of technical feasibility spikes, we dove into API documentation, experimented with connectors to support access to third-party data, and built prototypes to help Josh learn quickly and de-risk his product idea.
After the Discovery phase, we continued this work by building a proof-of-concept mobile app that Josh could test-drive and demo to partners and investors.
![Diagram showing the user flow for the Knect MVP.](https://images.prismic.io/thoughtbot-website/ZnG1Qpm069VX12oy_MVPflow.png?auto=format,compress)
Building product architecture and processing API data
This phase involved building a complex architecture in a very short amount of time.
Analyzing Knect users’ social and professional networks by tapping into various social platforms meant having lots of API requests and raw data.
Dealing with so many requests could result in very high costs, so we prioritized finding ways to process the data in a performant manner.
Our developers achieved this by using background jobs and mapping data into a common format that was usable in our app (instead of the raw API responses).
![Screenshots of the Knect proof-of-concept prototype.](https://images.prismic.io/thoughtbot-website/ZnFNApm069VX11PH_Grid.png?auto=format,compress)
Leveraging AI and heuristics
We wanted our proof-of-concept mobile app to provide useful recommendations for users to catch up and stay in touch with their network by only using a limited number of channels at first (connecting email messages and calendar events).
By experimenting with custom GPTs (using OpenAI’s ChatGPT 3.5 to identify relevant conversations) and writing complex heuristics in SQL to sift through data, we uncovered the first useful recommendations as we moved quickly through the last weeks of the project.
In parallel, we also made sure we were establishing a good base level of privacy and security, building a reliable app that could deal with so much sensitive information.
Product
“There was a lot of foundational work to be done to prove the feasibility of what we wanted to build. And obviously this research needed to be informed by the ultimate value proposition, because you don't want to spend resources building anything that you aren't actually going to use, that doesn’t provide real value to your users.” — Jordyn Bonds, Entrepreneur in Residence
Design
“During Discovery, gathering feedback and iterating on Knect’s value proposition as quickly as possible was fundamental to the success of the project. And prototyping was the fastest, most resource-effective tool for that purpose. In a matter of hours we could show our newest ideas to each other and to our users, learn what worked or what didn’t, and quickly design the next iteration.” — Caro Sotillo, Senior Product Designer
Engineering
“My priorities were making sure that we had a reliable secure app since we were dealing with such sensitive information. We had to think about security a lot, but from a user-facing standpoint we also had to make sure we were conveying valuable information in the right way.” — Diego Oliveira, Senior Mobile Developer
The outcome
Ready for the next steps
After the Discovery phase, we provided a research report gathering all the learnings about the problem, the market niche and the opportunity, a user-tested prototype, and a series of designs covering the ways in which the product could develop.
![Screenshot of the discovery report.](https://images.prismic.io/thoughtbot-website/ZnROJ5m069VX17Rk_Discovery.png?auto=format,compress)
![Screenshots of the Knect prototype, showing different recommendations to connect.](https://images.prismic.io/thoughtbot-website/Znl7q5bWFbowezoF_HorizScroll.png?auto=format,compress)
During the technical follow-up phase, we developed a proof-of-concept app that connected to the user's Google and LinkedIn accounts, analyzed the data, and provided recommendations of people the user should connect with based on a predetermined set of heuristics.
Knect founder Josh left with the confidence and tools to decide what the next steps of his venture would look like. Thanks to the upfront validation work, Josh discovered his idea for Knect was a valuable endeavor to pursue that would require full-time dedication, so he embarked on a journey to find other entrepreneurs ready to take on the challenge of moving Knect forward.
“Sometimes you need more than one smart person — we needed a team of the right people ready to dive in quickly. These senior engineers didn’t just tackle technical issues, they opened up a whole new way of looking at problems. It was fast, it was intense, and it totally changed the game for Knect.