HYbot
Northwestern University: MBA Kellogg Design Challenge 2024
Goal
Scaling AI Productivity for Product Management Teams
Role
Concept Strategist and Research Lead, Storytelling and Pitching
Tools
Human-Centered Design Research, Solution Design, Business Impact Pitch
Outcome
Our concept placed 7th out of 30+ teams
What is the KDC Design Challenge?
KDC 2024 Theme: AI x Design
The KDC design challenge, also known as the Kellogg Design Challenge, is a large-scale business design case competition hosted by Northwestern University's Kellogg School of Management, considered the world's biggest MBA design competition where teams of students tackle real-world business challenges presented by sponsoring companies, showcasing their design thinking and business acumen to develop actionable solutions.
Designing AI productivity solutions for product management and development teams.
KDC Challenge Prompt
How might one of the world leading global material handling solutions company leverage AI to cut product development time in half in a manner that is scalable across all global locations?
Solution: HYbot
A RAG-powered chatbot designed to seamlessly share and gather knowledge across the entire organization, tapping into the collective expertise of all team members and making valuable insights accessible to everyone in real time.
Step #1: Retrieve knowledge from a centralized HYG data pool
Step #2: Access context-rich data from a wide variety of company information
Step #3 and #4: Request data from higher management, teams, peers etc. and upload data you believe will benefit others in the company.
Step #5: Gamification is integrated into the tool to incentivize adoption and quality of data.
Our Solution Can Reduce Search Time by 25-45%
25%
Reduction in Search Time
(Healthcare)
20%
Faster Time to Market
(Healthcare)
45%
Reduction in Research Time
(Financial Services)
12%
Increase in Portfolio Returns
(Financial Services)
30%
Reduction in Search Costs
(Retail)
15%
Increase in Av. Order Value
(Healthcare)
Pitch Event
Interview Insights
Design Process
Number of Interviews: 15
Who: manufacturing engineers, software engineers, logistic and inventory managers, entrepreneurs, product leads
“(...) gathering the information can be one of the biggest pain points and slowdowns. You want to design something and have your skill set and your only roadblock to finishing it is time and information.”
(...) a lot of the time you're sitting and waiting for someone to reply to an email to figure out what they did on a test three years ago. It can be very time-consuming.”
- Jamie Palmer, Mechanical Engineer
Key Terms and Takeaways:
Continuous optimization
Total cost of ownership
Complexity of operations
Competitiveness of sector
Our team followed a double-diamond approach using mixed-methods research.
Social Media Ethnography
We Understood…
Most of the information needed exists. But is really hard to find.
It’s hard to know when someone needs knowledge, or has knowledge that could benefit others.
It’s hard to understand why decisions were made in the past.
The search for that knowledge is incredibly time consuming.
Team HMW
AI Model
Regular AI Chatbot VS RAG AI Chatbot
RAG can Overcome the Following Barriers:
Impacting the 6-Gate Product Development Process
G0
Strategy
G6
Strategy
Planning
Planning
The integration of RAG would greatly benefit the initial gates of the product development process as:
G0: Better understanding of the customer needs, market trends and tech innovations to identify opportunity zones
G1: Better informed outline and scopes of a project to confirm its viability.
G2: Better informed requirements that translate to fewer iterations, judgement calls, and ‘fill in the gaps’ in later stages.
Define
How might we enhance collaborative decision-making through up-to-date, easily accessible, verified knowledge to increase on-time deliveries?
G1
G2
G0
G1
Requirement
G2
Requirement
Using retrieval-augmented generation (RAG) to enable quick access to trusted, live, context-rich information, facilitating more accurate and relevant responses in real time while enhancing decision-making and problem-solving capabilities.
G3
Verification
G3
Validation
Verification
G4
G4
Validation
G5
Pre-product
G5
Closing
Pre-product
G6
Closing
With snowball effects on subsequent gates…..
Prototype I: Low-Fidelity
Implementation
-
4 Important Categories:
Cost of Components
Truck Ergonomics
Manufacturing Processes
Gearbox Layouts
-
Assign representatives within each ongoing project team (for QC)
Train them in using HYbot, and understanding the backend mechanics
-
Identify 3-5 ongoing low-risk projects for business sandbox testing prior to onboarding teams onto HYbot platform
-
Undertake user interviews and tech checks on the accuracy, speed, ripple effects downstream, etc.
-
Adding Categories:
Safety & Accidents
User Stories
Client Personas
Sustainability Metrics
-
Slack Integration
Email hybot-rag@hyg.com
Stack Overflow
Reflections
Through this project, I realized how crucial AI can be in alleviating the bottlenecks engineers often face when working without it. Without AI-driven tools to automate repetitive tasks or provide intelligent recommendations, engineers frequently struggle with waiting for manual inputs or guidance, which delays progress and hampers productivity. I learned that by integrating AI, not only can we streamline workflows and reduce waiting times, but we also empower engineers to focus on higher-value, creative problem-solving rather than getting bogged down in routine tasks.
Through RAG’s ability to streamline requirement gathering and reduce unnecessary iterations, I realized the importance of making data-backed decisions early to minimize risks and optimize the development process. This experience reinforced my belief in the value of AI and advanced technologies in driving efficiency and clarity throughout a project lifecycle.