SaySo_sakib

SaySo answers the question every Indian store owner asks: why did that customer walk out? A small badge clipped on the salesperson (XIAO ESP32-S3 in a 3D-printed case I designed) records the floor in one-minute pieces and uploads them over the store Wi-Fi. After closing, Meta's models turn the day into a short coaching report for the owner: which conversations won or lost a sale and what to say next time, complaints checked against what was really said, and one note per rep. The store's own CCTV can be added as an optional camera file. I field-tested it on 4 Nov 2026: I wore the badge as a shopper (with consent) in a footwear store. From 12 minutes of mixed Hindi and English audio, Muse Spark found on its own that the staff kept answering me in Hindi after I asked in English, and chatted among themselves about closing, and that the sale was lost. Nothing about the visit is written into the prompt. Customer voices are off by default: Spark picks out the customer's lines and the engine drops them, unless the customer agrees. Judges can test it live with no setup: open the field-test report, switch to Inspect mode to see every Meta model call (audio, transcripts, the exact Spark prompt and reply, timings), and press "Run the report again, live". Live site: https://sakibshaikh1507--sayso.ap-south.modal.run/dashboard/electronics/2026-11-04 Why it matters: in Indian retail, walkouts are invisible. Owners see only the till. SaySo gives them the moment, the words and the fix.