consulting
Slurrp Farm
AI powered business intelligence for a $90M consumer foods brand.
context
Wholsum Foods is the parent company of Slurrp Farm and Mille, founded by Oxbridge and Harvard alumni, selling packaged food for kids and families across ecommerce, quick commerce, and D2C in India, the UK, the US, and the UAE. I worked at Slurrp Farm early in my career; this engagement was pitched to the founders directly.
the thinking
Consumer signal at Wholsum Foods is scattered: reviews, search behaviour, social mentions, and competitor moves each sit in a different place and mostly get read reactively, if at all. The first build closed that gap with a single listening tool, but one tool trying to serve every job risks the opposite failure: information overload, where everyone reads the same dashboard through their own bias and nobody acts. The fix was not more data but fewer, sharper claims, and a separate tool for each real job: the founder listening to the market, the care team resolving it day to day, and the brand team testing a new idea before spending against it.
what i made
What started as a v0 sprint is now three tools on one shared corpus of 15,000-plus customer verbatims. Each one has a distinct job, and each refuses to do the others' job.
Mill
Social listening on steroids. Pulls what customers say about Slurrp Farm from every internal and external platform into one corpus, segments it, and surfaces the trends and insights inside it, every claim traceable to the sentence behind it.
Individual source counts held back at Wholsum's request.
Care
Turns the same data into an intelligence layer for the support team, drafting context-aware replies across email and WhatsApp so a conversation becomes retention, not just a closed ticket.
Kernel
An AI agent for go-to-market and new product hypothesis testing. The team brings a product idea, Kernel tests it against what real customers have actually said, and returns a sourced case for and against it, a GTM plan, and a confidence-rated verdict.
None of the three decides for you.
expected outcomes
modelled, not measured
Ranges I would underwrite, not results reported back to me.
Mill60-85% faster to produce a category read. 2-3x more analyst questions answerable per month.
Care50-70% less triage time per conversation. 20-40% more cases handled per agent.
Kernel60-85% faster first-pass hypothesis assessment. 2-3x more product ideas properly evaluated per planning cycle.