Kei AI assisant

Everyone has an AI agent. LeanDNA’s advantage: Kei is designed for supply chain professionals, not adapted from a generic chatbot. Ask detailed supply chain questions and Kei will deliver precise analysis with part numbers, quantities, suppliers, and root causes.

Company: LeanDNA

Role: Prinicipal Product Designer

Partners: Founder, Developer, Product management

Problem

LeanDNA’s APEX platform is trained on real supply chain concepts and complexity and generates supply chain recommendations. But users need to understand the reasoning behind these recommendations. A friendly ChatGPT-like interface wasn’t enough; Kei had to show its work. My goal was to build an AI assistant that explains its reasoning step-by-step, making users confident in Kei.

Iterations

• V1 | Proof of concept: Single conversation with no history
• V2 | Conversation management: Added conversation management. Users expected to see history like other ChatGTP and Gemini.
• V3 | Privacy: Controls for opting in or out. Users worried about saving sensitive supply chain conversations.
• V4 | Embed Kei: Inventory optimization platform was the first one to receive it. I wanted to test out styling the response (not just a block of text), and include important call outs. 

Next steps

Supply chain managers see “23% of orders are clear to build” and want to understand why. Instead of navigating away, they can open Kei on the page. Kei explains the metric, the trend, the calculation, and suggests the next question.

Workbench recommends "split and move out" and Kei explains that "the primary trigger for this recommendation is a massive disparity between current supply and actual demand." This gives buyers more trust and confidence in future recommendations by seeing the "math" behind curtain.