Participated in public demo 2603 with Mantri

The problem [00:00] Kunal built Mantri after watching a friend's Army supply business in Guwahati run entirely on informal WhatsApp communication, a sprawl of a dozen or more groups mixing clients, suppliers, and staff, where conflicting instructions between groups (one saying deliver to Mumbai, another saying Delhi) and buried messages regularly caused missed deliveries, escalations to the owner, and firefighting instead of business growth. What they built [02:45] Mantri monitors every WhatsApp conversation in real time, using a routing layer to map messages to entities, a detection agent to catch new orders, an update agent to extract supplier/item/quantity/date/commitment details, and a linkage agent to match client orders to supplier fulfillment. When a milestone is at risk, it escalates to staff automatically before the client has to ask. What happened [05:55] Evaluated end-to-end against 1,086 real client, supplier, and internal WhatsApp messages from an actual operating business (January–April 2026), Mantri scored 86.9/100 on a four-stage DeepEval pipeline: perfect routing accuracy and extraction quality, and agent-level LLM-judge scores of 0.86 (detection), 0.74 (update), and 0.90 (linkage), while surfacing 750+ escalations and ambiguities and discovering a new entity live in production. The ask [07:00] Kunal is looking for beta users, supply businesses running on WhatsApp, preferably based in India, along with introductions to anyone drowning in WhatsApp order management, and a co-founder with deep domain experience in supply chain or B2B sales in India.