MindsDB: An enterprise AI tool

Problem statement

MindsDB is a 0 to 1 product to connect and unify enterprise data; everything from databases, warehouses, files and websites enabling users to draw insights from across all their data. This is a non-trivial problem; architectures must have low latency, high accuracy, observability and great UX. The human aspects of adopting non-deterministic apps like this are also significant. Issues I faced were how to manage user expectations, onboarding, data quality and how the app should gracefully degrade.

Product management

I worked with the MindsDB product team to evaluate the market for this kind of tool and craft a go-to-market strategy. This included benchmarking, gathering and evaluating test datasets and writing PRDs and roadmaps. I worked closely with the sales team to identify design partners and then facilitated requirements gathering in conjunction with these partners. I then created actionable requirements for myself and the engineering teams.

Design strategy

• User research: The PM work segued nicely into user research to establish the target user personas. This included generative user research using interviews.com and Dovetail, facilitation with design partners using the Luma human-centered design methods and industry and academic research such as literature reviews.

  
  

• Information architecture: I designed a scalable and extensible IA that seemed like a viable fit with the target personas requirements and mental models. This included complicated workflows with multiple validation steps such as connecting to a datasource and evaluating the quality and permissions it contained. I used smart defaults, integrated tutorials, examples, screen tours and other onboarding strategies to guide new users.

  
  

• High-Fidelity Prototyping: I designed the UI using Tailwind, ShadCN and Figma. I prototyped key user journeys for usability testing. I used Voiceflow to prototype and evaluate the conversation design aspects of the chatbot.