Designing an onboarding experience for a AI product
Relevance AI is a no-code platform to analyze unstructured data
🔥 THE CHALLENGE
Previously, our onboarding experience received limited attention as we primarily relied on outreach sales. However, with a notable increase in self-serve users through word of mouth, recommendations, and SEO, it became crucial to convert them into paying users by providing a better onboarding experience.
Educating and guiding users on AI's capabilities
🌈 Solution design
To address this, we swiftly implemented a new flow and released it to production. Interestingly, we noticed an influx of users who enthusiastically attempted to resolve issues they encountered while uploading their datasets. This was likely because they recognized the value of our product through demos. This engagement was significant for us, as we had never experienced such proactive user involvement before. Additionally, we introduced sample datasets that users could utilize when encountering errors during the dataset upload process.
Below is the design in progress that was tested with users.
Design, build, validate and iterate!
To improve the onboarding experience, we acknowledged that encountering errors right after signing up was unpleasant for users. Through user interviews, we discovered that this issue was common, as our platform allowed AI analysis of unstructured data but required data reformatting. However, the main motivation for users to sign up was to see what our AI could do with their data. To address this, we decided to showcase quick demos during onboarding, demonstrating the capabilities of our AI. This aimed to clarify the value our product could provide and reduce drop-offs resulting from unclear expectations.
💎 Problem analysis
As a startup, we lacked statistically significant data volumes, so I relied heavily on qualitative feedback from our customers, team members, design reviews, and user session recordings. Through user research, we identified the following key issues:
-
Failed Dataset Uploads and Error Roadblocks: Users frequently encountered difficulties when attempting to upload their datasets, often being hindered by various errors.
-
Lack of Understanding of AI's Capabilities on User Datasets: Users expressed a lack of awareness regarding what AI could accomplish with their specific datasets. This lack of clarity contributed to their uncertainty about the potential value our product could provide.
-
Uncertainty about Post-Onboarding Actions: After completing the onboarding process, users were unsure about the subsequent steps to take. They lacked clear guidance on how to proceed and maximize the benefits of our platform.
The gif below shows how the previous experience was like:
Learning from different sources
After conducting additional rounds of user testing, we discovered ways to enhance this experience based on observations and user feedback.
-
To allow users to proceed to the next demo without having to close and open another demo modal, thus creating a more efficient flow.
-
To enable users to directly test their own data within the demos, eliminating the need to browse or search for workflows later.
-
To provide users with a better understanding of how credits can be utilized and why they are important to earn.
With these iterative improvements, we arrived at the following design.
.webp)
.webp)
.webp)
Also, you can see the new user flow in action below.
One valuable lesson I learned was that, even for technically oriented products, an educational onboarding experience can be beneficial for both users and our business.
Previously, I had designed an overly simplified onboarding process, assuming that no onboarding was the best approach and that simplicity was paramount. Fortunately, this time, I was able to enhance our product experience and grow as a designer.
There are numerous other aspects of the onboarding redesign that I cannot share here, but we remain committed to improving the experience through further user research, user session reviews, and ideation. 🔥
* Side note: as an extension of this project, I also designed the whole new website design including visuals, animation and implementation as you can see below. You can check it out at www.relevanceai.com
