Gen AI
KAI x BCG X (Knowledge.AI)
Gen AI INTERNAL ENTERPRISE TOOL | Lead Product Designer
4 Week MVP incubation
The Opportunity
Due to the large amounts of unorganized data across businesses, it is difficult to access or search data without spending exorbitant amounts of money on manually cleaning, organizing, and transferring the data into structured databases. KAI’s objective was to create an enterprise-grade, low-code tool that would allow Partners, Product Managers, Designers and Business Teams at all technical skill levels to utilize client’s unstructured proprietary data – pull out relevant data points – and, in collaboration with data scientists, establish, validate and sell use cases that would provide innovative client solutions.
Highlights
The Admin Portal serves as the main landing page where users can manage and create secure libraries
Once certain permissions are validated through SSO, the user lands on the KAI Admin Portal home. They are able to view owned libraries as well as other libraries they’ve been given access to and create new libraries.
The need for a ChatBot Playground was recognized as an immediate business priority
Given the rise in popularity of ChatBots (thanks ChatGPT) as well as the wide variety of use cases ChatBot’s can support, a playground for the KAI ChatBot was prioritized and developed for MVP.
The Approach
Due to our short development timeline, it was imperative to run two tracks of design deliverables to ensure we would hit critical milestones on time. A design system was delivered within days of the project kickoff in order to begin the front end development track of work while the prep, scheduling and administering of the research/discovery track happened simultaneously.
Several UX Artifacts were spun up to run ideation workshops in order to understand requirements that would support an Admin Portal and a ChatBot experience
These scrappy ideation workshops were also used to create alignment on key milestones within the various user journeys, ultimately shaping a baseline for Information Architecture and User Flows.
In parallel, ethnographic interviews were run to empathize with user needs and understand where our product could truly add value from day one
A discussion guide and card sort activity were utilized to conduct belief audits, understand pain points and ideate on what features would create the biggest impact out of the gate. This would ultimately influence the product roadmap.
As hypothesized, one of the key insights from our testing was how confusing the technical language of the experience was creating a feeling of anxiety and a potential risk of user abandonment
In order to burn down this risk, a “technical jargon” ideation workshop was set up with product, design, and engineering teams as well as a few of our end-users. This directly impacted the semantics and naming conventions referenced across the front end development and end-user experience.
Naming conventions were then added back into their appropriate IA trees in order to validate from a bird’s eye view
The Solution
KAI provides businesses the ability to customize their own proprietary Gen AI solution for a wide range of use cases, improving the overall profitability of the business as well as supporting a positive user experience. Built with privacy in mind – on a closed domain system – the product integrates with any natural language processing (NLP) model (e.g., Co:here, OpenAI, etc.). The ready-to-implement enterprise-grade tool enables any user to pilot a demonstration that highlights value and validates use-case hypotheses using best-in-class AI and data science.
The MVP Experience
LIBRARY CREATION & CHATBOT UI
From the Admin Portal Landing, the user creates a new library by adding new library details, selecting an available template and determining whether they’d like to configure the parameters of the template or go with default presets. They can then securely upload client data
The Library Detail page was built to adapt based on library output, displaying relevant information, such as analytics and annotation management for a ChatBot. It also acts as the main entry point to the output’s Playground
File management, template configuration details, user permissions and other key library features are easily accessed from the page as well.
The ChatBot experience includes onboarding to prompt writing as well as multi-chat management
The prompt onboarding experience was also built to be configurable so that it would be easy to update as we continued to learn from user knowledge gaps.
