Seven and a half years on the commercial side of CPG

I started as a Business Analyst and CPG domain consultant. I turned commercial problems into process maps, requirements and acceptance criteria, and I led the requirements for the first build of a global sales-execution platform: customer planning, outlet visits, order capture, promotions, contract compliance, and the CRM, ERP and logistics integrations behind them.

I also set up one shared requirements framework for delivery teams from several vendors, and assessed revenue growth management, digital commerce and loyalty options for European CPG companies.

I owned the mobile platform that replaced 56 legacy field-sales tools with one offline-first app, covering about 60% of the client's global revenue, with a 30+ person team across engineering, QA, architecture and change management.

Adoption stalled at 60%. The dashboards showed what reps used but couldn't explain why they avoided the rest, so I shadowed reps in the most resistant market. The global build had dropped functions some countries still needed. We kept one global process and built local needs in as country settings, and adoption rose to 70%.

Contracts also committed each outlet to a product mix and live promotions, but nobody at headquarters could see a shelf. I productized a third-party computer-vision model into a closed loop, and measured it on whether the gap closed after the rep acted.

I led product and go-to-market for an agentic revenue growth management (RGM) platform: software that helps large CPG companies decide price, promotion and product mix. Interviews with Revenue Management Directors and Commercial Finance leads at 8 target accounts said the same thing: they asked for help deciding what to do next, which moved the roadmap from reporting to recommendations.

Buyers kept stalling on data residency, vendor lock-in and "we already bought a model". I built the product on each customer's own LLM and data, which removed all three objections at once. And because a wrong autonomous price or promotion can breach a contract before anyone notices, the agent recommends and a person approves every consequential change.

After 7.5 years of shipping, I stepped back to ask what we had never built, and why. The answer became FieldIQ, an AI field-selling app for CPG reps that I built, from discovery and prompt design to test suites and integration.

Four agents work in the live prototype. Draft Order Agent proposes an order line by line, each with a reason and the numbers behind it, and the rep edits before submitting. CommCheck sorts pre-visit signals into Act Today, Opportunity and Escalate. Territory Pulse writes a ten-second territory summary. OnboardIQ finds where a new outlet is stuck.

Every release has to pass a test suite. On the seeded run the order agent passed 22 of 26 cases with 3 critical misses, so the gate failed and blocked that release.